Add GPT-6.2 Sol and Astra to model catalogs (#62106)

## What changed

- Add `gpt-6.2-sol` and `gpt-6.2-astra` to the bundled, Amazon Bedrock, and Vertex catalogs, including Bedrock Runtime global and US cross-region variants.
- Make GPT-6.2 Sol the default catalog selection and refresh model ordering, descriptions, and bundled instructions.
- Increase the GPT-6.1 Sol context window to 500,000 tokens.
- Advertise the `ultrafast` service tier for GPT-6.2 Astra on Bedrock.
- Supply the same current model catalog to both executors in the prompt-prefix compatibility test so catalog updates do not cause prefix mismatches.

## Testing

Extend provider, app-server, and TUI coverage for catalog ordering, model fallback, Vertex requests, and GPT-6.2 Astra's Bedrock `ultrafast` selection and clearing behavior. Refresh model-picker and request snapshots.

GitOrigin-RevId: f28940e4a255fcdd88f7114873259bd1f3399812
This commit is contained in:
andrewgu-oai
2026-11-17 20:04:36 +00:00
committed by copyberry
parent 9a56c2b1aa
commit 6c0cee502d
95 changed files with 1286 additions and 737 deletions
+25 -13
View File
@@ -1,5 +1,5 @@
//! Compare model-visible prefixes from two app-server versions using the same client and inputs.
//! The contract runner supplies a released CLI; ordinary runs use the current binary twice.
//! Compare model-visible prefixes from two app-server versions using the same catalog and inputs.
//! The contract runner supplies a released CLI; ordinary runs check same-version determinism.
//! Resume coverage checks that changed tool declarations leave the retained prefix intact.
#[path = "prompt_prefix/resume_tests.rs"]
@@ -23,6 +23,14 @@ use tempfile::TempDir;
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn prompt_prefix_matches_release() -> Result<()> {
let release = std::env::var_os("CONTRACT_EXECUTOR_BIN").map(std::path::PathBuf::from);
// Catalog content is an input to the harness. Give both executors the same
// current catalog so normal model and prompt updates cannot cause skew.
let catalog_dir = TempDir::new()?;
let catalog_path = catalog_dir.path().join("models.json");
std::fs::write(
&catalog_path,
serde_json::to_vec(&codex_models_manager::bundled_models_response()?)?,
)?;
let mut prefixes = Vec::new();
for program in [release.as_deref(), None] {
@@ -37,10 +45,12 @@ async fn prompt_prefix_matches_release() -> Result<()> {
)
.await;
let home = TempDir::new()?;
// Use each binary's bundled model catalog and feature defaults. Supplying a
// catalog or base instructions from the test would hide changes to them.
MockResponsesConfig::new(&server.uri())
.with_model("gpt-6-astra")
.with_root_config(&format!(
"model_catalog_json = {}",
serde_json::to_string(&catalog_path)?
))
// CCA disables bundled skills in its orchestrator.
.with_root_config("skills.bundled.enabled = false")
.write(home.path())?;
@@ -124,7 +134,7 @@ async fn prompt_prefix_matches_release() -> Result<()> {
}
assert_eq!(
prefixes[0], prefixes[1],
"the default inference prefix differs from the released version; gate the change behind a default-off feature key"
"the inference prefix differs from the released version for identical catalog and inputs; gate the harness change behind a default-off feature key"
);
Ok(())
}
@@ -154,15 +164,17 @@ fn accept_spawn_agent_shortlist_rollout(released: &mut Value, current: &Value) {
return;
};
const OLD_MODELS: &str = r#"Available model overrides (optional; inherited parent model is preferred):
- `gpt-6.1-sol`: Latest workhorse model for coding and everyday work. Reasoning efforts: low (default), medium, high, xhigh, max, ultra. Service tiers: priority.
- `gpt-6-astra`: Frontier intelligence for the most demanding work. Reasoning efforts: low (default), medium, high, xhigh, max, ultra. Service tiers: priority.
- `gpt-6-sol`: Previous generation workhorse model. Reasoning efforts: low, medium (default), high, xhigh, max, ultra. Service tiers: priority.
- `gpt-6-luna`: Fast and affordable model for easier tasks. Reasoning efforts: low, medium (default), high, xhigh, max. Service tiers: priority.
- `gpt-5.6-sol`: Older generation workhorse model. Reasoning efforts: low (default), medium, high, xhigh, max, ultra. Service tiers: priority.
- `gpt-6.2-sol`: Fast, go-to model for coding and everyday work. Reasoning efforts: low (default), medium, high, xhigh, max, ultra. Service tiers: priority.
- `gpt-6.2-astra`: Our most capable model for complex coding and professional work. Reasoning efforts: low (default), medium, high, xhigh, max, ultra. Service tiers: priority.
- `gpt-6.1-terra`: Efficient model that excels at visual understanding and computer use. Reasoning efforts: low (default), medium, high, xhigh, max, ultra. Service tiers: priority.
- `gpt-6.1-sol`: Previous generation model for coding and everyday work. Reasoning efforts: low (default), medium, high, xhigh, max, ultra. Service tiers: priority.
- `gpt-6-astra`: Previous generation model for complex coding and professional work. Reasoning efforts: low (default), medium, high, xhigh, max, ultra. Service tiers: priority.
"#;
const EXTRA_MODELS: &str = r#"- `gpt-6.1-terra`: GPT-6.1 Terra Codex model. Reasoning efforts: low (default), medium, high, xhigh, max, ultra. Service tiers: priority.
- `gpt-5.6-terra`: Older balanced model for straightforward work. Reasoning efforts: low, medium (default), high, xhigh, max, ultra. Service tiers: priority.
- `gpt-5.6-luna`: Older fast and efficient model. Reasoning efforts: low, medium (default), high, xhigh, max. Service tiers: priority.
const EXTRA_MODELS: &str = r#"- `gpt-6-sol`: Legacy model for coding and everyday work. Reasoning efforts: low, medium (default), high, xhigh, max, ultra. Service tiers: priority.
- `gpt-6-luna`: Fast and affordable model for easier tasks. Reasoning efforts: low, medium (default), high, xhigh, max. Service tiers: priority.
- `gpt-5.6-sol`: Legacy model for coding and everyday work. Reasoning efforts: low (default), medium, high, xhigh, max, ultra. Service tiers: priority.
- `gpt-5.6-terra`: Legacy balanced model for agentic coding. Reasoning efforts: low, medium (default), high, xhigh, max, ultra. Service tiers: priority.
- `gpt-5.6-luna`: Legacy fast model for simpler tasks. Reasoning efforts: low, medium (default), high, xhigh, max. Service tiers: priority.
- `gpt-5.5`: Legacy coding model. Reasoning efforts: low, medium (default), high, xhigh. Service tiers: priority.
"#;
if old.matches(OLD_MODELS).count() != 1 {
@@ -8,8 +8,8 @@ use codex_app_server_protocol::ModelServiceTier;
use pretty_assertions::assert_eq;
use tempfile::TempDir;
#[test_case::test_case("amazon-bedrock", vec!["openai.gpt-6.1-sol", "openai.gpt-6-astra"]; "mantle")]
#[test_case::test_case("amazon-bedrock-runtime", vec!["global.openai.gpt-6.1-sol", "global.openai.gpt-6-astra", "us.openai.gpt-6.1-sol", "us.openai.gpt-6-astra"]; "runtime")]
#[test_case::test_case("amazon-bedrock", vec!["openai.gpt-6.2-astra", "openai.gpt-6.1-sol", "openai.gpt-6-astra"]; "mantle")]
#[test_case::test_case("amazon-bedrock-runtime", vec!["global.openai.gpt-6.2-astra", "global.openai.gpt-6.1-sol", "global.openai.gpt-6-astra", "us.openai.gpt-6.2-astra", "us.openai.gpt-6.1-sol", "us.openai.gpt-6-astra"]; "runtime")]
#[tokio::test]
async fn bedrock_model_list_advertises_ultrafast_without_changing_default(
provider_id: &str,
@@ -115,7 +115,7 @@ model = "gpt-5.4-mini"
})
.await?;
assert_eq!(response.model, "openai.gpt-6.1-sol");
assert_eq!(response.model, "openai.gpt-6.2-sol");
Ok(())
}
@@ -246,6 +246,8 @@ async fn thread_start_provider_model_fallback_uses_bedrock_static_catalog() -> R
.await?;
for model in [
"openai.gpt-6.1-sol",
"openai.gpt-6.2-astra",
"openai.gpt-6.2-sol",
"openai.gpt-6.1-terra",
"openai.gpt-6-sol",
"openai.gpt-6-luna",
@@ -275,7 +277,7 @@ async fn thread_start_provider_model_fallback_uses_bedrock_static_catalog() -> R
supported_with_fallback.model,
unsupported_without_fallback.model,
],
vec!["openai.gpt-6.1-sol", "openai.gpt-5.5", "gpt-5.4-mini"]
vec!["openai.gpt-6.2-sol", "openai.gpt-5.5", "gpt-5.4-mini"]
);
Ok(())
}
@@ -296,6 +298,10 @@ async fn thread_start_bedrock_runtime_prefers_global_cross_region_models() -> Re
for model in [
"global.openai.gpt-6.1-sol",
"us.openai.gpt-6.1-sol",
"global.openai.gpt-6.2-astra",
"us.openai.gpt-6.2-astra",
"global.openai.gpt-6.2-sol",
"us.openai.gpt-6.2-sol",
"global.openai.gpt-6.1-terra",
"us.openai.gpt-6.1-terra",
"global.openai.gpt-6-sol",
@@ -317,7 +323,7 @@ async fn thread_start_bedrock_runtime_prefers_global_cross_region_models() -> Re
/*allow_provider_model_fallback*/ true,
)
.await?;
assert_eq!(response.model, "global.openai.gpt-6.1-sol");
assert_eq!(response.model, "global.openai.gpt-6.2-sol");
Ok(())
}
+2
View File
@@ -96,6 +96,8 @@ async fn vertex_credential_file_is_retained_when_refreshing_after_401() -> Resul
#[test_case("gpt-6.1-sol", false; "sol")]
#[test_case("gpt-6.1-terra", false; "terra")]
#[test_case("gpt-6.2-astra", false; "astra_62")]
#[test_case("gpt-6.2-sol", false; "sol_62")]
#[test_case("gpt-6.1-sol", true; "sol_incremental")]
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn vertex_requests_use_command_auth_with_namespaces_and_tool_search(
@@ -3,10 +3,25 @@ source: core/tests/suite/gcp.rs
expression: "models.iter().map(|model|\n(&model.model, model.show_in_picker, model.is_default)).collect::<Vec<_>>()"
---
[
(
"gpt-6.2-sol",
true,
true,
),
(
"gpt-6.2-astra",
true,
false,
),
(
"gpt-6.1-terra",
true,
false,
),
(
"gpt-6.1-sol",
true,
true,
false,
),
(
"gpt-6-astra",
@@ -18,9 +33,4 @@ expression: "models.iter().map(|model|\n(&model.model, model.show_in_picker, mod
true,
false,
),
(
"gpt-6.1-terra",
true,
false,
),
]
@@ -15,7 +15,7 @@ Settings:
stream: true
text: {"verbosity":"low"}
tool_choice: "auto"
tools (11; hash=9f125ad8b5fb8fda):
tools (11; hash=df07401579d70e22):
- function/exec_command: Runs a command in a PTY, returning output or a session ID for ongoing interaction.; args=[cmd, justification, login, max_output_tokens, prefix_rule, sandbox_permissions, shell, tty, workdir, yield_time_ms]; hash=24f4b6739dd5d940
- function/write_stdin: Writes characters to an existing unified exec session and returns recent output.; args=[chars, max_output_tokens, session_id, yield_time_ms]; hash=7fb62d8a3e2df4e2
- function/list_mcp_resources: Lists resources provided by MCP servers. Resources allow servers to share data that provides cont...; args=[cursor, server]; hash=6792edaff13561bd
@@ -24,7 +24,7 @@ Settings:
- function/request_user_input: Request user input for one to three short questions and wait for the response. This tool is only ...; args=[questions]; hash=64a729b25c424fc6
- custom/apply_patch: The `apply_patch` tool can be used to edit files. This is a FREEFORM tool, so do not wrap the pat...; hash=dea9f12b0111a4ba
- function/view_image: View a local image file from the filesystem when visual inspection is needed. Use this for images...; args=[path]; hash=1908de9c0be7118a
- namespace/multi_agent_v1: Tools for spawning and managing sub-agents.; hash=141ddefee46d1d08
- namespace/multi_agent_v1: Tools for spawning and managing sub-agents.; hash=2da9b5fbea45fffe
- function/close_agent
- function/resume_agent
- function/send_input
@@ -15,7 +15,7 @@ Settings:
stream: true
text: {"verbosity":"low"}
tool_choice: "auto"
tools (11; hash=9f125ad8b5fb8fda):
tools (11; hash=df07401579d70e22):
- function/exec_command: Runs a command in a PTY, returning output or a session ID for ongoing interaction.; args=[cmd, justification, login, max_output_tokens, prefix_rule, sandbox_permissions, shell, tty, workdir, yield_time_ms]; hash=24f4b6739dd5d940
- function/write_stdin: Writes characters to an existing unified exec session and returns recent output.; args=[chars, max_output_tokens, session_id, yield_time_ms]; hash=7fb62d8a3e2df4e2
- function/list_mcp_resources: Lists resources provided by MCP servers. Resources allow servers to share data that provides cont...; args=[cursor, server]; hash=6792edaff13561bd
@@ -24,7 +24,7 @@ Settings:
- function/request_user_input: Request user input for one to three short questions and wait for the response. This tool is only ...; args=[questions]; hash=64a729b25c424fc6
- custom/apply_patch: The `apply_patch` tool can be used to edit files. This is a FREEFORM tool, so do not wrap the pat...; hash=dea9f12b0111a4ba
- function/view_image: View a local image file from the filesystem when visual inspection is needed. Use this for images...; args=[path]; hash=1908de9c0be7118a
- namespace/multi_agent_v1: Tools for spawning and managing sub-agents.; hash=141ddefee46d1d08
- namespace/multi_agent_v1: Tools for spawning and managing sub-agents.; hash=2da9b5fbea45fffe
- function/close_agent
- function/resume_agent
- function/send_input
@@ -15,7 +15,7 @@ Settings:
stream: true
text: {"verbosity":"low"}
tool_choice: "auto"
tools (11; hash=9f125ad8b5fb8fda):
tools (11; hash=df07401579d70e22):
- function/exec_command: Runs a command in a PTY, returning output or a session ID for ongoing interaction.; args=[cmd, justification, login, max_output_tokens, prefix_rule, sandbox_permissions, shell, tty, workdir, yield_time_ms]; hash=24f4b6739dd5d940
- function/write_stdin: Writes characters to an existing unified exec session and returns recent output.; args=[chars, max_output_tokens, session_id, yield_time_ms]; hash=7fb62d8a3e2df4e2
- function/list_mcp_resources: Lists resources provided by MCP servers. Resources allow servers to share data that provides cont...; args=[cursor, server]; hash=6792edaff13561bd
@@ -24,7 +24,7 @@ Settings:
- function/request_user_input: Request user input for one to three short questions and wait for the response. This tool is only ...; args=[questions]; hash=64a729b25c424fc6
- custom/apply_patch: The `apply_patch` tool can be used to edit files. This is a FREEFORM tool, so do not wrap the pat...; hash=dea9f12b0111a4ba
- function/view_image: View a local image file from the filesystem when visual inspection is needed. Use this for images...; args=[path]; hash=1908de9c0be7118a
- namespace/multi_agent_v1: Tools for spawning and managing sub-agents.; hash=141ddefee46d1d08
- namespace/multi_agent_v1: Tools for spawning and managing sub-agents.; hash=2da9b5fbea45fffe
- function/close_agent
- function/resume_agent
- function/send_input
@@ -15,7 +15,7 @@ Settings:
stream: true
text: {"verbosity":"low"}
tool_choice: "auto"
tools (11; hash=9f125ad8b5fb8fda):
tools (11; hash=df07401579d70e22):
- function/exec_command: Runs a command in a PTY, returning output or a session ID for ongoing interaction.; args=[cmd, justification, login, max_output_tokens, prefix_rule, sandbox_permissions, shell, tty, workdir, yield_time_ms]; hash=24f4b6739dd5d940
- function/write_stdin: Writes characters to an existing unified exec session and returns recent output.; args=[chars, max_output_tokens, session_id, yield_time_ms]; hash=7fb62d8a3e2df4e2
- function/list_mcp_resources: Lists resources provided by MCP servers. Resources allow servers to share data that provides cont...; args=[cursor, server]; hash=6792edaff13561bd
@@ -24,7 +24,7 @@ Settings:
- function/request_user_input: Request user input for one to three short questions and wait for the response. This tool is only ...; args=[questions]; hash=64a729b25c424fc6
- custom/apply_patch: The `apply_patch` tool can be used to edit files. This is a FREEFORM tool, so do not wrap the pat...; hash=dea9f12b0111a4ba
- function/view_image: View a local image file from the filesystem when visual inspection is needed. Use this for images...; args=[path]; hash=1908de9c0be7118a
- namespace/multi_agent_v1: Tools for spawning and managing sub-agents.; hash=141ddefee46d1d08
- namespace/multi_agent_v1: Tools for spawning and managing sub-agents.; hash=2da9b5fbea45fffe
- function/close_agent
- function/resume_agent
- function/send_input
@@ -15,7 +15,7 @@ Settings:
stream: true
text: {"verbosity":"low"}
tool_choice: "auto"
tools (11; hash=9f125ad8b5fb8fda):
tools (11; hash=df07401579d70e22):
- function/exec_command: Runs a command in a PTY, returning output or a session ID for ongoing interaction.; args=[cmd, justification, login, max_output_tokens, prefix_rule, sandbox_permissions, shell, tty, workdir, yield_time_ms]; hash=24f4b6739dd5d940
- function/write_stdin: Writes characters to an existing unified exec session and returns recent output.; args=[chars, max_output_tokens, session_id, yield_time_ms]; hash=7fb62d8a3e2df4e2
- function/list_mcp_resources: Lists resources provided by MCP servers. Resources allow servers to share data that provides cont...; args=[cursor, server]; hash=6792edaff13561bd
@@ -24,7 +24,7 @@ Settings:
- function/request_user_input: Request user input for one to three short questions and wait for the response. This tool is only ...; args=[questions]; hash=64a729b25c424fc6
- custom/apply_patch: The `apply_patch` tool can be used to edit files. This is a FREEFORM tool, so do not wrap the pat...; hash=dea9f12b0111a4ba
- function/view_image: View a local image file from the filesystem when visual inspection is needed. Use this for images...; args=[path]; hash=1908de9c0be7118a
- namespace/multi_agent_v1: Tools for spawning and managing sub-agents.; hash=141ddefee46d1d08
- namespace/multi_agent_v1: Tools for spawning and managing sub-agents.; hash=2da9b5fbea45fffe
- function/close_agent
- function/resume_agent
- function/send_input
@@ -15,7 +15,7 @@ Settings:
stream: true
text: {"verbosity":"low"}
tool_choice: "auto"
tools (11; hash=9f125ad8b5fb8fda):
tools (11; hash=df07401579d70e22):
- function/exec_command: Runs a command in a PTY, returning output or a session ID for ongoing interaction.; args=[cmd, justification, login, max_output_tokens, prefix_rule, sandbox_permissions, shell, tty, workdir, yield_time_ms]; hash=24f4b6739dd5d940
- function/write_stdin: Writes characters to an existing unified exec session and returns recent output.; args=[chars, max_output_tokens, session_id, yield_time_ms]; hash=7fb62d8a3e2df4e2
- function/list_mcp_resources: Lists resources provided by MCP servers. Resources allow servers to share data that provides cont...; args=[cursor, server]; hash=6792edaff13561bd
@@ -24,7 +24,7 @@ Settings:
- function/request_user_input: Request user input for one to three short questions and wait for the response. This tool is only ...; args=[questions]; hash=64a729b25c424fc6
- custom/apply_patch: The `apply_patch` tool can be used to edit files. This is a FREEFORM tool, so do not wrap the pat...; hash=dea9f12b0111a4ba
- function/view_image: View a local image file from the filesystem when visual inspection is needed. Use this for images...; args=[path]; hash=1908de9c0be7118a
- namespace/multi_agent_v1: Tools for spawning and managing sub-agents.; hash=141ddefee46d1d08
- namespace/multi_agent_v1: Tools for spawning and managing sub-agents.; hash=2da9b5fbea45fffe
- function/close_agent
- function/resume_agent
- function/send_input
@@ -15,7 +15,7 @@ Settings:
stream: true
text: {"verbosity":"low"}
tool_choice: "auto"
tools (11; hash=9f125ad8b5fb8fda):
tools (11; hash=df07401579d70e22):
- function/exec_command: Runs a command in a PTY, returning output or a session ID for ongoing interaction.; args=[cmd, justification, login, max_output_tokens, prefix_rule, sandbox_permissions, shell, tty, workdir, yield_time_ms]; hash=24f4b6739dd5d940
- function/write_stdin: Writes characters to an existing unified exec session and returns recent output.; args=[chars, max_output_tokens, session_id, yield_time_ms]; hash=7fb62d8a3e2df4e2
- function/list_mcp_resources: Lists resources provided by MCP servers. Resources allow servers to share data that provides cont...; args=[cursor, server]; hash=6792edaff13561bd
@@ -24,7 +24,7 @@ Settings:
- function/request_user_input: Request user input for one to three short questions and wait for the response. This tool is only ...; args=[questions]; hash=64a729b25c424fc6
- custom/apply_patch: The `apply_patch` tool can be used to edit files. This is a FREEFORM tool, so do not wrap the pat...; hash=dea9f12b0111a4ba
- function/view_image: View a local image file from the filesystem when visual inspection is needed. Use this for images...; args=[path]; hash=1908de9c0be7118a
- namespace/multi_agent_v1: Tools for spawning and managing sub-agents.; hash=141ddefee46d1d08
- namespace/multi_agent_v1: Tools for spawning and managing sub-agents.; hash=2da9b5fbea45fffe
- function/close_agent
- function/resume_agent
- function/send_input
@@ -1,18 +1,17 @@
---
source: core/tests/suite/scenarios_agent_eviction_tests.rs
assertion_line: 275
expression: "format!(\"{history}\\n\\nResumed worker mailbox:\\n{mailbox}\")"
---
Scenario: A parent with no configured environment delegates an API migration, queues a note while the worker is idle, and follows up after idle eviction. The worker temporarily attaches a workspace, then resumes with no environment, its earlier answer, and the unread note. Complete model streams are grouped as parent, then worker.
## Window 1
-- request 1 (turn) --
00:additional_tools/developer (2; hash=59b9daa63f2c8258):
00:additional_tools/developer (2; hash=1efa2dc9d797cace):
- namespace/functions; hash=f33f647fee88e25e
- custom/exec
- function/wait
- function/request_user_input
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=0271b89d55152d99
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=a097b9364047b8bb
- function/followup_task
- function/interrupt_agent
- function/list_agents
@@ -66,12 +65,12 @@ Scenario: A parent with no configured environment delegates an API migration, qu
## Window 2 (after request 6: input diverged at item 03)
-- request 7 (turn) --
00:additional_tools/developer (2; hash=59b9daa63f2c8258):
00:additional_tools/developer (2; hash=1efa2dc9d797cace):
- namespace/functions; hash=f33f647fee88e25e
- custom/exec
- function/wait
- function/request_user_input
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=0271b89d55152d99
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=a097b9364047b8bb
- function/followup_task
- function/interrupt_agent
- function/list_agents
@@ -6,7 +6,7 @@ Scenario: Permission discovery preserves legacy output, exposes reply-only acces
## Window 1
-- request 1 (request) --
00:additional_tools/developer (3; hash=7004828fd9d4bdda):
00:additional_tools/developer (3; hash=c7c5759e6d0f0042):
- namespace/functions; hash=c39a9fd756ad9f5b
- function/exec_command
- function/write_stdin
@@ -17,7 +17,7 @@ Scenario: Permission discovery preserves legacy output, exposes reply-only acces
- namespace/clock: Tools for reading and waiting on time.; hash=a80435c7a3a3943f
- function/curr_time
- function/sleep
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=2171bf8582b06f98
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=1e271dc249fbf952
- function/create_channel
- function/followup_task
- function/get_channels
@@ -6,7 +6,7 @@ Scenario: Astra sees two environments with different permissions. Only the secon
## Window 1
-- request 1 (turn) --
00:additional_tools/developer (3; hash=f388ebd7783b858c):
00:additional_tools/developer (3; hash=7aafbac510cab352):
- namespace/functions; hash=ebc92fe078f78ade
- custom/exec
- function/wait
@@ -14,7 +14,7 @@ Scenario: Astra sees two environments with different permissions. Only the secon
- function/request_user_input_async
- namespace/clock: Tools for reading and waiting on time.; hash=f402e8c5e9b5e317
- function/sleep
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=0271b89d55152d99
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=a097b9364047b8bb
- function/followup_task
- function/interrupt_agent
- function/list_agents
@@ -30,14 +30,14 @@ Scenario: Astra sees two environments with different permissions. Only the secon
User authorization and preferences persist across turns. Do not request permission again when the user h... any guidelines provided in skills or external files. [hash=911f91c92c72882c]
<OMITTED 155 LINES; ~4935 TOKENS; hash=5573869716192298>
<OMITTED 151 LINES; ~4851 TOKENS; hash=959762226fedfd0c>
## Decision Boundary
- Skill naming: If a plugin contributes skills, those skill entries are prefixed with plugin_name: in the Skills list.
- MCP naming: Plugin-provided MCP tools keep standard MCP identifiers such as mcp__server__tool; use tool provenance to tell which plugin they come from.
- Trigger rules: If the user explicitly names a plugin, prefer capabilities associated with that plugin for that turn.
- Relationship to capabilities: Plugins are not invoked directly. Use their underlying skills, MCP tools, and app tools to help solve the task.
- Relevance: Determine what a plugin can help with from explicit user mention or from the plugin-associa..., MCP tools, and apps exposed elsewhere in this turn. [hash=620459a4e7b590f0]
- Missing/blocked: If the user requests a plugin that does not have relevant callable capabilities for t..., say so briefly and continue with the best fallback. [hash=746b0469a8f6cb3e]
Discover and use relevant tools proactively when they may help carry out the requested task. For example...ng the user for information they may already contain. [hash=3af5c93cbc00b4c7]
When a plugin is available for a service and its tools are sufficient for completing the task, prefer th...er has an explicit preference, defer to that instead. [hash=936dbd7fe76c210f]
A user can also tag a plugin using `[@plugin-name](plugin://{{plugin_id}})` or an app using `[$app-name]... app, prefer its relevant capabilities for that turn. [hash=8730f69f41534f8f]
02:message/developer[3]:
[01] <permissions instructions><environment id="environment-a">
@@ -6,7 +6,7 @@ Scenario: Astra compacts a completed launch-notes turn in the background. The us
## Window 1
-- request 1 (request) --
00:additional_tools/developer (3; hash=f2e4edd02b6f9300):
00:additional_tools/developer (3; hash=a7c46a0959d54866):
- namespace/functions; hash=81846de5bc4e7474
- custom/exec
- function/wait
@@ -14,7 +14,7 @@ Scenario: Astra compacts a completed launch-notes turn in the background. The us
- function/request_user_input_async
- namespace/clock: Tools for reading and waiting on time.; hash=f402e8c5e9b5e317
- function/sleep
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=0271b89d55152d99
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=a097b9364047b8bb
- function/followup_task
- function/interrupt_agent
- function/list_agents
@@ -45,7 +45,7 @@ Scenario: Astra compacts a completed launch-notes turn in the background. The us
## Window 2 (after request 2: input diverged at item 10)
-- request 3 (request) --
00:additional_tools/developer (3; hash=f2e4edd02b6f9300):
00:additional_tools/developer (3; hash=a7c46a0959d54866):
- namespace/functions; hash=81846de5bc4e7474
- custom/exec
- function/wait
@@ -53,7 +53,7 @@ Scenario: Astra compacts a completed launch-notes turn in the background. The us
- function/request_user_input_async
- namespace/clock: Tools for reading and waiting on time.; hash=f402e8c5e9b5e317
- function/sleep
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=0271b89d55152d99
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=a097b9364047b8bb
- function/followup_task
- function/interrupt_agent
- function/list_agents
@@ -88,7 +88,7 @@ Scenario: Astra compacts a completed launch-notes turn in the background. The us
## Window 3 (after request 3: input diverged at item 02)
-- request 4 (request) --
00:additional_tools/developer (3; hash=f2e4edd02b6f9300):
00:additional_tools/developer (3; hash=a7c46a0959d54866):
- namespace/functions; hash=81846de5bc4e7474
- custom/exec
- function/wait
@@ -96,7 +96,7 @@ Scenario: Astra compacts a completed launch-notes turn in the background. The us
- function/request_user_input_async
- namespace/clock: Tools for reading and waiting on time.; hash=f402e8c5e9b5e317
- function/sleep
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=0271b89d55152d99
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=a097b9364047b8bb
- function/followup_task
- function/interrupt_agent
- function/list_agents
@@ -6,7 +6,7 @@ Scenario: Astra uses the catalog question schema to ask who a launch update is f
## Window 1
-- request 1 (request) --
00:additional_tools/developer (3; hash=a17d5f400f9e670a):
00:additional_tools/developer (3; hash=1575ec6449565e60):
- namespace/functions; hash=1f321c5169dfcdc0
- custom/exec
- function/wait
@@ -14,7 +14,7 @@ Scenario: Astra uses the catalog question schema to ask who a launch update is f
- function/request_user_input_async
- namespace/clock: Tools for reading and waiting on time.; hash=f402e8c5e9b5e317
- function/sleep
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=0271b89d55152d99
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=a097b9364047b8bb
- function/followup_task
- function/interrupt_agent
- function/list_agents
@@ -6,7 +6,7 @@ Scenario: Astra reads code-mode timings and the collector notice even at zero to
## Window 1
-- request 1 (turn) --
00:additional_tools/developer (3; hash=f2e4edd02b6f9300):
00:additional_tools/developer (3; hash=a7c46a0959d54866):
- namespace/functions; hash=81846de5bc4e7474
- custom/exec
- function/wait
@@ -14,7 +14,7 @@ Scenario: Astra reads code-mode timings and the collector notice even at zero to
- function/request_user_input_async
- namespace/clock: Tools for reading and waiting on time.; hash=f402e8c5e9b5e317
- function/sleep
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=0271b89d55152d99
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=a097b9364047b8bb
- function/followup_task
- function/interrupt_agent
- function/list_agents
@@ -30,14 +30,14 @@ Scenario: Astra reads code-mode timings and the collector notice even at zero to
User authorization and preferences persist across turns. Do not request permission again when the user h... any guidelines provided in skills or external files. [hash=911f91c92c72882c]
<OMITTED 155 LINES; ~4935 TOKENS; hash=5573869716192298>
<OMITTED 151 LINES; ~4851 TOKENS; hash=959762226fedfd0c>
## Decision Boundary
- Skill naming: If a plugin contributes skills, those skill entries are prefixed with plugin_name: in the Skills list.
- MCP naming: Plugin-provided MCP tools keep standard MCP identifiers such as mcp__server__tool; use tool provenance to tell which plugin they come from.
- Trigger rules: If the user explicitly names a plugin, prefer capabilities associated with that plugin for that turn.
- Relationship to capabilities: Plugins are not invoked directly. Use their underlying skills, MCP tools, and app tools to help solve the task.
- Relevance: Determine what a plugin can help with from explicit user mention or from the plugin-associa..., MCP tools, and apps exposed elsewhere in this turn. [hash=620459a4e7b590f0]
- Missing/blocked: If the user requests a plugin that does not have relevant callable capabilities for t..., say so briefly and continue with the best fallback. [hash=746b0469a8f6cb3e]
Discover and use relevant tools proactively when they may help carry out the requested task. For example...ng the user for information they may already contain. [hash=3af5c93cbc00b4c7]
When a plugin is available for a service and its tools are sufficient for completing the task, prefer th...er has an explicit preference, defer to that instead. [hash=936dbd7fe76c210f]
A user can also tag a plugin using `[@plugin-name](plugin://{{plugin_id}})` or an app using `[$app-name]... app, prefer its relevant capabilities for that turn. [hash=8730f69f41534f8f]
02:message/developer[2]:
[01] <permissions instructions>
@@ -16,7 +16,7 @@ Settings:
text: {"verbosity":"low"}
tool_choice: "auto"
-- request 1 (request) --
00:additional_tools/developer (3; hash=aa1c8d2cbd56c247):
00:additional_tools/developer (3; hash=39cb3ee73f2e1987):
- namespace/functions; hash=c4874773c24012ab
- custom/exec
- function/wait
@@ -24,7 +24,7 @@ Settings:
- function/request_user_input_async
- namespace/clock: Tools for reading and waiting on time.; hash=f402e8c5e9b5e317
- function/sleep
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=0271b89d55152d99
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=a097b9364047b8bb
- function/followup_task
- function/interrupt_agent
- function/list_agents
@@ -40,14 +40,14 @@ Settings:
User authorization and preferences persist across turns. Do not request permission again when the user h... any guidelines provided in skills or external files. [hash=911f91c92c72882c]
<OMITTED 155 LINES; ~4935 TOKENS; hash=5573869716192298>
<OMITTED 151 LINES; ~4851 TOKENS; hash=959762226fedfd0c>
## Decision Boundary
- Skill naming: If a plugin contributes skills, those skill entries are prefixed with plugin_name: in the Skills list.
- MCP naming: Plugin-provided MCP tools keep standard MCP identifiers such as mcp__server__tool; use tool provenance to tell which plugin they come from.
- Trigger rules: If the user explicitly names a plugin, prefer capabilities associated with that plugin for that turn.
- Relationship to capabilities: Plugins are not invoked directly. Use their underlying skills, MCP tools, and app tools to help solve the task.
- Relevance: Determine what a plugin can help with from explicit user mention or from the plugin-associa..., MCP tools, and apps exposed elsewhere in this turn. [hash=620459a4e7b590f0]
- Missing/blocked: If the user requests a plugin that does not have relevant callable capabilities for t..., say so briefly and continue with the best fallback. [hash=746b0469a8f6cb3e]
Discover and use relevant tools proactively when they may help carry out the requested task. For example...ng the user for information they may already contain. [hash=3af5c93cbc00b4c7]
When a plugin is available for a service and its tools are sufficient for completing the task, prefer th...er has an explicit preference, defer to that instead. [hash=936dbd7fe76c210f]
A user can also tag a plugin using `[@plugin-name](plugin://{{plugin_id}})` or an app using `[$app-name]... app, prefer its relevant capabilities for that turn. [hash=8730f69f41534f8f]
02:message/developer[3]:
[01] <permissions instructions>
@@ -16,7 +16,7 @@ Settings:
text: {"verbosity":"low"}
tool_choice: "auto"
-- request 1 (request) --
00:additional_tools/developer (3; hash=aa1c8d2cbd56c247):
00:additional_tools/developer (3; hash=39cb3ee73f2e1987):
- namespace/functions; hash=c4874773c24012ab
- custom/exec
- function/wait
@@ -24,7 +24,7 @@ Settings:
- function/request_user_input_async
- namespace/clock: Tools for reading and waiting on time.; hash=f402e8c5e9b5e317
- function/sleep
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=0271b89d55152d99
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=a097b9364047b8bb
- function/followup_task
- function/interrupt_agent
- function/list_agents
@@ -40,14 +40,14 @@ Settings:
User authorization and preferences persist across turns. Do not request permission again when the user h... any guidelines provided in skills or external files. [hash=911f91c92c72882c]
<OMITTED 155 LINES; ~4935 TOKENS; hash=5573869716192298>
<OMITTED 151 LINES; ~4851 TOKENS; hash=959762226fedfd0c>
## Decision Boundary
- Skill naming: If a plugin contributes skills, those skill entries are prefixed with plugin_name: in the Skills list.
- MCP naming: Plugin-provided MCP tools keep standard MCP identifiers such as mcp__server__tool; use tool provenance to tell which plugin they come from.
- Trigger rules: If the user explicitly names a plugin, prefer capabilities associated with that plugin for that turn.
- Relationship to capabilities: Plugins are not invoked directly. Use their underlying skills, MCP tools, and app tools to help solve the task.
- Relevance: Determine what a plugin can help with from explicit user mention or from the plugin-associa..., MCP tools, and apps exposed elsewhere in this turn. [hash=620459a4e7b590f0]
- Missing/blocked: If the user requests a plugin that does not have relevant callable capabilities for t..., say so briefly and continue with the best fallback. [hash=746b0469a8f6cb3e]
Discover and use relevant tools proactively when they may help carry out the requested task. For example...ng the user for information they may already contain. [hash=3af5c93cbc00b4c7]
When a plugin is available for a service and its tools are sufficient for completing the task, prefer th...er has an explicit preference, defer to that instead. [hash=936dbd7fe76c210f]
A user can also tag a plugin using `[@plugin-name](plugin://{{plugin_id}})` or an app using `[$app-name]... app, prefer its relevant capabilities for that turn. [hash=8730f69f41534f8f]
02:message/developer[3]:
[01] <permissions instructions>
@@ -6,7 +6,7 @@ Scenario: Astra receives new user input after its unfinished response is interru
## Window 1
-- request 1 (request) --
00:additional_tools/developer (3; hash=f2e4edd02b6f9300):
00:additional_tools/developer (3; hash=a7c46a0959d54866):
- namespace/functions; hash=81846de5bc4e7474
- custom/exec
- function/wait
@@ -14,7 +14,7 @@ Scenario: Astra receives new user input after its unfinished response is interru
- function/request_user_input_async
- namespace/clock: Tools for reading and waiting on time.; hash=f402e8c5e9b5e317
- function/sleep
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=0271b89d55152d99
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=a097b9364047b8bb
- function/followup_task
- function/interrupt_agent
- function/list_agents
@@ -6,7 +6,7 @@ Scenario: Astra continues after new user input yields a running code cell, then
## Window 1
-- request 1 (turn) --
00:additional_tools/developer (3; hash=836c706f54159a7d):
00:additional_tools/developer (3; hash=6f23592af0082595):
- namespace/functions; hash=d55357e79b53812d
- custom/exec
- function/wait
@@ -14,7 +14,7 @@ Scenario: Astra continues after new user input yields a running code cell, then
- function/request_user_input_async
- namespace/clock: Tools for reading and waiting on time.; hash=f402e8c5e9b5e317
- function/sleep
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=0271b89d55152d99
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=a097b9364047b8bb
- function/followup_task
- function/interrupt_agent
- function/list_agents
@@ -30,14 +30,14 @@ Scenario: Astra continues after new user input yields a running code cell, then
User authorization and preferences persist across turns. Do not request permission again when the user h... any guidelines provided in skills or external files. [hash=911f91c92c72882c]
<OMITTED 155 LINES; ~4935 TOKENS; hash=5573869716192298>
<OMITTED 151 LINES; ~4851 TOKENS; hash=959762226fedfd0c>
## Decision Boundary
- Skill naming: If a plugin contributes skills, those skill entries are prefixed with plugin_name: in the Skills list.
- MCP naming: Plugin-provided MCP tools keep standard MCP identifiers such as mcp__server__tool; use tool provenance to tell which plugin they come from.
- Trigger rules: If the user explicitly names a plugin, prefer capabilities associated with that plugin for that turn.
- Relationship to capabilities: Plugins are not invoked directly. Use their underlying skills, MCP tools, and app tools to help solve the task.
- Relevance: Determine what a plugin can help with from explicit user mention or from the plugin-associa..., MCP tools, and apps exposed elsewhere in this turn. [hash=620459a4e7b590f0]
- Missing/blocked: If the user requests a plugin that does not have relevant callable capabilities for t..., say so briefly and continue with the best fallback. [hash=746b0469a8f6cb3e]
Discover and use relevant tools proactively when they may help carry out the requested task. For example...ng the user for information they may already contain. [hash=3af5c93cbc00b4c7]
When a plugin is available for a service and its tools are sufficient for completing the task, prefer th...er has an explicit preference, defer to that instead. [hash=936dbd7fe76c210f]
A user can also tag a plugin using `[@plugin-name](plugin://{{plugin_id}})` or an app using `[$app-name]... app, prefer its relevant capabilities for that turn. [hash=8730f69f41534f8f]
02:message/developer[2]:
[01] <permissions instructions>
@@ -16,7 +16,7 @@ Settings:
text: {"verbosity":"low"}
tool_choice: "auto"
-- request 1 (turn) --
00:additional_tools/developer (3; hash=f2e4edd02b6f9300):
00:additional_tools/developer (3; hash=a7c46a0959d54866):
- namespace/functions; hash=81846de5bc4e7474
- custom/exec
- function/wait
@@ -24,7 +24,7 @@ Settings:
- function/request_user_input_async
- namespace/clock: Tools for reading and waiting on time.; hash=f402e8c5e9b5e317
- function/sleep
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=0271b89d55152d99
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=a097b9364047b8bb
- function/followup_task
- function/interrupt_agent
- function/list_agents
@@ -40,14 +40,14 @@ Settings:
User authorization and preferences persist across turns. Do not request permission again when the user h... any guidelines provided in skills or external files. [hash=911f91c92c72882c]
<OMITTED 155 LINES; ~4935 TOKENS; hash=5573869716192298>
<OMITTED 151 LINES; ~4851 TOKENS; hash=959762226fedfd0c>
## Decision Boundary
- Skill naming: If a plugin contributes skills, those skill entries are prefixed with plugin_name: in the Skills list.
- MCP naming: Plugin-provided MCP tools keep standard MCP identifiers such as mcp__server__tool; use tool provenance to tell which plugin they come from.
- Trigger rules: If the user explicitly names a plugin, prefer capabilities associated with that plugin for that turn.
- Relationship to capabilities: Plugins are not invoked directly. Use their underlying skills, MCP tools, and app tools to help solve the task.
- Relevance: Determine what a plugin can help with from explicit user mention or from the plugin-associa..., MCP tools, and apps exposed elsewhere in this turn. [hash=620459a4e7b590f0]
- Missing/blocked: If the user requests a plugin that does not have relevant callable capabilities for t..., say so briefly and continue with the best fallback. [hash=746b0469a8f6cb3e]
Discover and use relevant tools proactively when they may help carry out the requested task. For example...ng the user for information they may already contain. [hash=3af5c93cbc00b4c7]
When a plugin is available for a service and its tools are sufficient for completing the task, prefer th...er has an explicit preference, defer to that instead. [hash=936dbd7fe76c210f]
A user can also tag a plugin using `[@plugin-name](plugin://{{plugin_id}})` or an app using `[$app-name]... app, prefer its relevant capabilities for that turn. [hash=8730f69f41534f8f]
02:message/developer[3]:
[01] <skills_instructions>
@@ -188,7 +188,7 @@ Settings:
## Window 2 (after request 3: input diverged at item 02)
Settings: same as window 1
-- request 4 (turn) --
00:additional_tools/developer (3; hash=f2e4edd02b6f9300):
00:additional_tools/developer (3; hash=a7c46a0959d54866):
- namespace/functions; hash=81846de5bc4e7474
- custom/exec
- function/wait
@@ -196,7 +196,7 @@ Settings: same as window 1
- function/request_user_input_async
- namespace/clock: Tools for reading and waiting on time.; hash=f402e8c5e9b5e317
- function/sleep
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=0271b89d55152d99
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=a097b9364047b8bb
- function/followup_task
- function/interrupt_agent
- function/list_agents
@@ -212,14 +212,14 @@ Settings: same as window 1
User authorization and preferences persist across turns. Do not request permission again when the user h... any guidelines provided in skills or external files. [hash=911f91c92c72882c]
<OMITTED 155 LINES; ~4935 TOKENS; hash=5573869716192298>
<OMITTED 151 LINES; ~4851 TOKENS; hash=959762226fedfd0c>
## Decision Boundary
- Skill naming: If a plugin contributes skills, those skill entries are prefixed with plugin_name: in the Skills list.
- MCP naming: Plugin-provided MCP tools keep standard MCP identifiers such as mcp__server__tool; use tool provenance to tell which plugin they come from.
- Trigger rules: If the user explicitly names a plugin, prefer capabilities associated with that plugin for that turn.
- Relationship to capabilities: Plugins are not invoked directly. Use their underlying skills, MCP tools, and app tools to help solve the task.
- Relevance: Determine what a plugin can help with from explicit user mention or from the plugin-associa..., MCP tools, and apps exposed elsewhere in this turn. [hash=620459a4e7b590f0]
- Missing/blocked: If the user requests a plugin that does not have relevant callable capabilities for t..., say so briefly and continue with the best fallback. [hash=746b0469a8f6cb3e]
Discover and use relevant tools proactively when they may help carry out the requested task. For example...ng the user for information they may already contain. [hash=3af5c93cbc00b4c7]
When a plugin is available for a service and its tools are sufficient for completing the task, prefer th...er has an explicit preference, defer to that instead. [hash=936dbd7fe76c210f]
A user can also tag a plugin using `[@plugin-name](plugin://{{plugin_id}})` or an app using `[$app-name]... app, prefer its relevant capabilities for that turn. [hash=8730f69f41534f8f]
02:message/user:
Plan a team kickoff for Friday using $outline and $calendar:agenda.
@@ -16,7 +16,7 @@ Settings:
text: {"verbosity":"low"}
tool_choice: "auto"
-- request 1 (turn) --
00:additional_tools/developer (3; hash=f2e4edd02b6f9300):
00:additional_tools/developer (3; hash=a7c46a0959d54866):
- namespace/functions; hash=81846de5bc4e7474
- custom/exec
- function/wait
@@ -24,7 +24,7 @@ Settings:
- function/request_user_input_async
- namespace/clock: Tools for reading and waiting on time.; hash=f402e8c5e9b5e317
- function/sleep
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=0271b89d55152d99
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=a097b9364047b8bb
- function/followup_task
- function/interrupt_agent
- function/list_agents
@@ -40,14 +40,14 @@ Settings:
User authorization and preferences persist across turns. Do not request permission again when the user h... any guidelines provided in skills or external files. [hash=911f91c92c72882c]
<OMITTED 155 LINES; ~4935 TOKENS; hash=5573869716192298>
<OMITTED 151 LINES; ~4851 TOKENS; hash=959762226fedfd0c>
## Decision Boundary
- Skill naming: If a plugin contributes skills, those skill entries are prefixed with plugin_name: in the Skills list.
- MCP naming: Plugin-provided MCP tools keep standard MCP identifiers such as mcp__server__tool; use tool provenance to tell which plugin they come from.
- Trigger rules: If the user explicitly names a plugin, prefer capabilities associated with that plugin for that turn.
- Relationship to capabilities: Plugins are not invoked directly. Use their underlying skills, MCP tools, and app tools to help solve the task.
- Relevance: Determine what a plugin can help with from explicit user mention or from the plugin-associa..., MCP tools, and apps exposed elsewhere in this turn. [hash=620459a4e7b590f0]
- Missing/blocked: If the user requests a plugin that does not have relevant callable capabilities for t..., say so briefly and continue with the best fallback. [hash=746b0469a8f6cb3e]
Discover and use relevant tools proactively when they may help carry out the requested task. For example...ng the user for information they may already contain. [hash=3af5c93cbc00b4c7]
When a plugin is available for a service and its tools are sufficient for completing the task, prefer th...er has an explicit preference, defer to that instead. [hash=936dbd7fe76c210f]
A user can also tag a plugin using `[@plugin-name](plugin://{{plugin_id}})` or an app using `[$app-name]... app, prefer its relevant capabilities for that turn. [hash=8730f69f41534f8f]
02:message/developer[3]:
[01] <skills_instructions>
@@ -16,7 +16,7 @@ Settings:
text: {"verbosity":"low"}
tool_choice: "auto"
-- request 1 (turn) --
00:additional_tools/developer (3; hash=f2e4edd02b6f9300):
00:additional_tools/developer (3; hash=a7c46a0959d54866):
- namespace/functions; hash=81846de5bc4e7474
- custom/exec
- function/wait
@@ -24,7 +24,7 @@ Settings:
- function/request_user_input_async
- namespace/clock: Tools for reading and waiting on time.; hash=f402e8c5e9b5e317
- function/sleep
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=0271b89d55152d99
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=a097b9364047b8bb
- function/followup_task
- function/interrupt_agent
- function/list_agents
@@ -40,14 +40,14 @@ Settings:
User authorization and preferences persist across turns. Do not request permission again when the user h... any guidelines provided in skills or external files. [hash=911f91c92c72882c]
<OMITTED 155 LINES; ~4935 TOKENS; hash=5573869716192298>
<OMITTED 151 LINES; ~4851 TOKENS; hash=959762226fedfd0c>
## Decision Boundary
- Skill naming: If a plugin contributes skills, those skill entries are prefixed with plugin_name: in the Skills list.
- MCP naming: Plugin-provided MCP tools keep standard MCP identifiers such as mcp__server__tool; use tool provenance to tell which plugin they come from.
- Trigger rules: If the user explicitly names a plugin, prefer capabilities associated with that plugin for that turn.
- Relationship to capabilities: Plugins are not invoked directly. Use their underlying skills, MCP tools, and app tools to help solve the task.
- Relevance: Determine what a plugin can help with from explicit user mention or from the plugin-associa..., MCP tools, and apps exposed elsewhere in this turn. [hash=620459a4e7b590f0]
- Missing/blocked: If the user requests a plugin that does not have relevant callable capabilities for t..., say so briefly and continue with the best fallback. [hash=746b0469a8f6cb3e]
Discover and use relevant tools proactively when they may help carry out the requested task. For example...ng the user for information they may already contain. [hash=3af5c93cbc00b4c7]
When a plugin is available for a service and its tools are sufficient for completing the task, prefer th...er has an explicit preference, defer to that instead. [hash=936dbd7fe76c210f]
A user can also tag a plugin using `[@plugin-name](plugin://{{plugin_id}})` or an app using `[$app-name]... app, prefer its relevant capabilities for that turn. [hash=8730f69f41534f8f]
02:message/developer[3]:
[01] <skills_instructions>
@@ -16,7 +16,7 @@ Settings:
text: {"verbosity":"low"}
tool_choice: "auto"
-- request 1 (turn) --
00:additional_tools/developer (3; hash=f2e4edd02b6f9300):
00:additional_tools/developer (3; hash=a7c46a0959d54866):
- namespace/functions; hash=81846de5bc4e7474
- custom/exec
- function/wait
@@ -24,7 +24,7 @@ Settings:
- function/request_user_input_async
- namespace/clock: Tools for reading and waiting on time.; hash=f402e8c5e9b5e317
- function/sleep
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=0271b89d55152d99
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=a097b9364047b8bb
- function/followup_task
- function/interrupt_agent
- function/list_agents
@@ -40,14 +40,14 @@ Settings:
User authorization and preferences persist across turns. Do not request permission again when the user h... any guidelines provided in skills or external files. [hash=911f91c92c72882c]
<OMITTED 155 LINES; ~4935 TOKENS; hash=5573869716192298>
<OMITTED 151 LINES; ~4851 TOKENS; hash=959762226fedfd0c>
## Decision Boundary
- Skill naming: If a plugin contributes skills, those skill entries are prefixed with plugin_name: in the Skills list.
- MCP naming: Plugin-provided MCP tools keep standard MCP identifiers such as mcp__server__tool; use tool provenance to tell which plugin they come from.
- Trigger rules: If the user explicitly names a plugin, prefer capabilities associated with that plugin for that turn.
- Relationship to capabilities: Plugins are not invoked directly. Use their underlying skills, MCP tools, and app tools to help solve the task.
- Relevance: Determine what a plugin can help with from explicit user mention or from the plugin-associa..., MCP tools, and apps exposed elsewhere in this turn. [hash=620459a4e7b590f0]
- Missing/blocked: If the user requests a plugin that does not have relevant callable capabilities for t..., say so briefly and continue with the best fallback. [hash=746b0469a8f6cb3e]
Discover and use relevant tools proactively when they may help carry out the requested task. For example...ng the user for information they may already contain. [hash=3af5c93cbc00b4c7]
When a plugin is available for a service and its tools are sufficient for completing the task, prefer th...er has an explicit preference, defer to that instead. [hash=936dbd7fe76c210f]
A user can also tag a plugin using `[@plugin-name](plugin://{{plugin_id}})` or an app using `[$app-name]... app, prefer its relevant capabilities for that turn. [hash=8730f69f41534f8f]
02:message/developer[2]:
[01] <permissions instructions>
@@ -16,7 +16,7 @@ Settings:
text: {"verbosity":"low"}
tool_choice: "auto"
-- request 1 (turn) --
00:additional_tools/developer (3; hash=f2e4edd02b6f9300):
00:additional_tools/developer (3; hash=a7c46a0959d54866):
- namespace/functions; hash=81846de5bc4e7474
- custom/exec
- function/wait
@@ -24,7 +24,7 @@ Settings:
- function/request_user_input_async
- namespace/clock: Tools for reading and waiting on time.; hash=f402e8c5e9b5e317
- function/sleep
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=0271b89d55152d99
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=a097b9364047b8bb
- function/followup_task
- function/interrupt_agent
- function/list_agents
@@ -40,14 +40,14 @@ Settings:
User authorization and preferences persist across turns. Do not request permission again when the user h... any guidelines provided in skills or external files. [hash=911f91c92c72882c]
<OMITTED 155 LINES; ~4935 TOKENS; hash=5573869716192298>
<OMITTED 151 LINES; ~4851 TOKENS; hash=959762226fedfd0c>
## Decision Boundary
- Skill naming: If a plugin contributes skills, those skill entries are prefixed with plugin_name: in the Skills list.
- MCP naming: Plugin-provided MCP tools keep standard MCP identifiers such as mcp__server__tool; use tool provenance to tell which plugin they come from.
- Trigger rules: If the user explicitly names a plugin, prefer capabilities associated with that plugin for that turn.
- Relationship to capabilities: Plugins are not invoked directly. Use their underlying skills, MCP tools, and app tools to help solve the task.
- Relevance: Determine what a plugin can help with from explicit user mention or from the plugin-associa..., MCP tools, and apps exposed elsewhere in this turn. [hash=620459a4e7b590f0]
- Missing/blocked: If the user requests a plugin that does not have relevant callable capabilities for t..., say so briefly and continue with the best fallback. [hash=746b0469a8f6cb3e]
Discover and use relevant tools proactively when they may help carry out the requested task. For example...ng the user for information they may already contain. [hash=3af5c93cbc00b4c7]
When a plugin is available for a service and its tools are sufficient for completing the task, prefer th...er has an explicit preference, defer to that instead. [hash=936dbd7fe76c210f]
A user can also tag a plugin using `[@plugin-name](plugin://{{plugin_id}})` or an app using `[$app-name]... app, prefer its relevant capabilities for that turn. [hash=8730f69f41534f8f]
02:message/developer[2]:
[01] <permissions instructions>
@@ -16,7 +16,7 @@ Settings:
text: {"verbosity":"low"}
tool_choice: "auto"
-- request 1 (turn) --
00:additional_tools/developer (3; hash=f2e4edd02b6f9300):
00:additional_tools/developer (3; hash=a7c46a0959d54866):
- namespace/functions; hash=81846de5bc4e7474
- custom/exec
- function/wait
@@ -24,7 +24,7 @@ Settings:
- function/request_user_input_async
- namespace/clock: Tools for reading and waiting on time.; hash=f402e8c5e9b5e317
- function/sleep
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=0271b89d55152d99
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=a097b9364047b8bb
- function/followup_task
- function/interrupt_agent
- function/list_agents
@@ -40,14 +40,14 @@ Settings:
User authorization and preferences persist across turns. Do not request permission again when the user h... any guidelines provided in skills or external files. [hash=911f91c92c72882c]
<OMITTED 155 LINES; ~4935 TOKENS; hash=5573869716192298>
<OMITTED 151 LINES; ~4851 TOKENS; hash=959762226fedfd0c>
## Decision Boundary
- Skill naming: If a plugin contributes skills, those skill entries are prefixed with plugin_name: in the Skills list.
- MCP naming: Plugin-provided MCP tools keep standard MCP identifiers such as mcp__server__tool; use tool provenance to tell which plugin they come from.
- Trigger rules: If the user explicitly names a plugin, prefer capabilities associated with that plugin for that turn.
- Relationship to capabilities: Plugins are not invoked directly. Use their underlying skills, MCP tools, and app tools to help solve the task.
- Relevance: Determine what a plugin can help with from explicit user mention or from the plugin-associa..., MCP tools, and apps exposed elsewhere in this turn. [hash=620459a4e7b590f0]
- Missing/blocked: If the user requests a plugin that does not have relevant callable capabilities for t..., say so briefly and continue with the best fallback. [hash=746b0469a8f6cb3e]
Discover and use relevant tools proactively when they may help carry out the requested task. For example...ng the user for information they may already contain. [hash=3af5c93cbc00b4c7]
When a plugin is available for a service and its tools are sufficient for completing the task, prefer th...er has an explicit preference, defer to that instead. [hash=936dbd7fe76c210f]
A user can also tag a plugin using `[@plugin-name](plugin://{{plugin_id}})` or an app using `[$app-name]... app, prefer its relevant capabilities for that turn. [hash=8730f69f41534f8f]
02:message/developer[3]:
[01] <permissions instructions>
@@ -6,7 +6,7 @@ Scenario: A side chat receives the parent turn as reference, adds its own turn,
## Window 1
-- request 1 (turn) --
00:additional_tools/developer (3; hash=f2e4edd02b6f9300):
00:additional_tools/developer (3; hash=a7c46a0959d54866):
- namespace/functions; hash=81846de5bc4e7474
- custom/exec
- function/wait
@@ -14,7 +14,7 @@ Scenario: A side chat receives the parent turn as reference, adds its own turn,
- function/request_user_input_async
- namespace/clock: Tools for reading and waiting on time.; hash=f402e8c5e9b5e317
- function/sleep
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=0271b89d55152d99
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=a097b9364047b8bb
- function/followup_task
- function/interrupt_agent
- function/list_agents
@@ -30,14 +30,14 @@ Scenario: A side chat receives the parent turn as reference, adds its own turn,
User authorization and preferences persist across turns. Do not request permission again when the user h... any guidelines provided in skills or external files. [hash=911f91c92c72882c]
<OMITTED 155 LINES; ~4935 TOKENS; hash=5573869716192298>
<OMITTED 151 LINES; ~4851 TOKENS; hash=959762226fedfd0c>
## Decision Boundary
- Skill naming: If a plugin contributes skills, those skill entries are prefixed with plugin_name: in the Skills list.
- MCP naming: Plugin-provided MCP tools keep standard MCP identifiers such as mcp__server__tool; use tool provenance to tell which plugin they come from.
- Trigger rules: If the user explicitly names a plugin, prefer capabilities associated with that plugin for that turn.
- Relationship to capabilities: Plugins are not invoked directly. Use their underlying skills, MCP tools, and app tools to help solve the task.
- Relevance: Determine what a plugin can help with from explicit user mention or from the plugin-associa..., MCP tools, and apps exposed elsewhere in this turn. [hash=620459a4e7b590f0]
- Missing/blocked: If the user requests a plugin that does not have relevant callable capabilities for t..., say so briefly and continue with the best fallback. [hash=746b0469a8f6cb3e]
Discover and use relevant tools proactively when they may help carry out the requested task. For example...ng the user for information they may already contain. [hash=3af5c93cbc00b4c7]
When a plugin is available for a service and its tools are sufficient for completing the task, prefer th...er has an explicit preference, defer to that instead. [hash=936dbd7fe76c210f]
A user can also tag a plugin using `[@plugin-name](plugin://{{plugin_id}})` or an app using `[$app-name]... app, prefer its relevant capabilities for that turn. [hash=8730f69f41534f8f]
02:message/developer[3]:
[01] Use plain language.
@@ -108,7 +108,7 @@ Scenario: A side chat receives the parent turn as reference, adds its own turn,
## Window 2 (after request 1: settings changed (prompt_cache_key))
-- request 2 (turn) --
00:additional_tools/developer (3; hash=f2e4edd02b6f9300):
00:additional_tools/developer (3; hash=a7c46a0959d54866):
- namespace/functions; hash=81846de5bc4e7474
- custom/exec
- function/wait
@@ -116,7 +116,7 @@ Scenario: A side chat receives the parent turn as reference, adds its own turn,
- function/request_user_input_async
- namespace/clock: Tools for reading and waiting on time.; hash=f402e8c5e9b5e317
- function/sleep
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=0271b89d55152d99
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=a097b9364047b8bb
- function/followup_task
- function/interrupt_agent
- function/list_agents
@@ -132,14 +132,14 @@ Scenario: A side chat receives the parent turn as reference, adds its own turn,
User authorization and preferences persist across turns. Do not request permission again when the user h... any guidelines provided in skills or external files. [hash=911f91c92c72882c]
<OMITTED 155 LINES; ~4935 TOKENS; hash=5573869716192298>
<OMITTED 151 LINES; ~4851 TOKENS; hash=959762226fedfd0c>
## Decision Boundary
- Skill naming: If a plugin contributes skills, those skill entries are prefixed with plugin_name: in the Skills list.
- MCP naming: Plugin-provided MCP tools keep standard MCP identifiers such as mcp__server__tool; use tool provenance to tell which plugin they come from.
- Trigger rules: If the user explicitly names a plugin, prefer capabilities associated with that plugin for that turn.
- Relationship to capabilities: Plugins are not invoked directly. Use their underlying skills, MCP tools, and app tools to help solve the task.
- Relevance: Determine what a plugin can help with from explicit user mention or from the plugin-associa..., MCP tools, and apps exposed elsewhere in this turn. [hash=620459a4e7b590f0]
- Missing/blocked: If the user requests a plugin that does not have relevant callable capabilities for t..., say so briefly and continue with the best fallback. [hash=746b0469a8f6cb3e]
Discover and use relevant tools proactively when they may help carry out the requested task. For example...ng the user for information they may already contain. [hash=3af5c93cbc00b4c7]
When a plugin is available for a service and its tools are sufficient for completing the task, prefer th...er has an explicit preference, defer to that instead. [hash=936dbd7fe76c210f]
A user can also tag a plugin using `[@plugin-name](plugin://{{plugin_id}})` or an app using `[$app-name]... app, prefer its relevant capabilities for that turn. [hash=8730f69f41534f8f]
02:message/developer[3]:
[01] Use plain language.
@@ -6,7 +6,7 @@ Scenario: Core renders the host's roster with native World State cadence. Compac
## Window 1
-- request 1 (turn) --
00:additional_tools/developer (3; hash=f2e4edd02b6f9300):
00:additional_tools/developer (3; hash=a7c46a0959d54866):
- namespace/functions; hash=81846de5bc4e7474
- custom/exec
- function/wait
@@ -14,7 +14,7 @@ Scenario: Core renders the host's roster with native World State cadence. Compac
- function/request_user_input_async
- namespace/clock: Tools for reading and waiting on time.; hash=f402e8c5e9b5e317
- function/sleep
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=0271b89d55152d99
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=a097b9364047b8bb
- function/followup_task
- function/interrupt_agent
- function/list_agents
@@ -63,7 +63,7 @@ Scenario: Core renders the host's roster with native World State cadence. Compac
## Window 2 (after request 4: input diverged at item 02)
-- request 5 (turn) --
00:additional_tools/developer (3; hash=f2e4edd02b6f9300):
00:additional_tools/developer (3; hash=a7c46a0959d54866):
- namespace/functions; hash=81846de5bc4e7474
- custom/exec
- function/wait
@@ -71,7 +71,7 @@ Scenario: Core renders the host's roster with native World State cadence. Compac
- function/request_user_input_async
- namespace/clock: Tools for reading and waiting on time.; hash=f402e8c5e9b5e317
- function/sleep
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=0271b89d55152d99
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=a097b9364047b8bb
- function/followup_task
- function/interrupt_agent
- function/list_agents
@@ -14,7 +14,7 @@ Settings:
store: false
stream: true
tool_choice: "auto"
tools (11; hash=2b4c921518a4db6f):
tools (11; hash=067d271a36f2eb01):
- function/exec_command: Runs a command in a PTY, returning output or a session ID for ongoing interaction.; args=[cmd, justification, login, max_output_tokens, prefix_rule, sandbox_permissions, shell, tty, workdir, yield_time_ms]; hash=24f4b6739dd5d940
- function/write_stdin: Writes characters to an existing unified exec session and returns recent output.; args=[chars, max_output_tokens, session_id, yield_time_ms]; hash=7fb62d8a3e2df4e2
- function/list_mcp_resources: Lists resources provided by MCP servers. Resources allow servers to share data that provides cont...; args=[cursor, server]; hash=6792edaff13561bd
@@ -23,7 +23,7 @@ Settings:
- function/update_plan: Updates the task plan. Provide an optional explanation and a list of plan items, each with a step...; args=[explanation, plan]; hash=cbba5ed302a9c0eb
- function/request_user_input: Request user input for one to three short questions and wait for the response. This tool is only ...; args=[questions]; hash=64a729b25c424fc6
- function/view_image: View a local image file from the filesystem when visual inspection is needed. Use this for images...; args=[path]; hash=1908de9c0be7118a
- namespace/multi_agent_v1: Tools for spawning and managing sub-agents.; hash=f066a006a1fd060e
- namespace/multi_agent_v1: Tools for spawning and managing sub-agents.; hash=c041c3bd4f63167c
- function/close_agent
- function/resume_agent
- function/send_input
@@ -15,7 +15,7 @@ Settings:
stream: true
text: {"verbosity":"low"}
tool_choice: "auto"
tools (11; hash=d2f0f4f212de42c4):
tools (11; hash=2d4682354915df6c):
- function/exec_command: Runs a command in a PTY, returning output or a session ID for ongoing interaction.; args=[cmd, justification, login, max_output_tokens, prefix_rule, sandbox_permissions, shell, tty, workdir, yield_time_ms]; hash=24f4b6739dd5d940
- function/write_stdin: Writes characters to an existing unified exec session and returns recent output.; args=[chars, max_output_tokens, session_id, yield_time_ms]; hash=7fb62d8a3e2df4e2
- function/list_mcp_resources: Lists resources provided by MCP servers. Resources allow servers to share data that provides cont...; args=[cursor, server]; hash=6792edaff13561bd
@@ -24,7 +24,7 @@ Settings:
- function/request_user_input: Request user input for one to three short questions and wait for the response. This tool is only ...; args=[questions]; hash=64a729b25c424fc6
- custom/apply_patch: The `apply_patch` tool can be used to edit files. This is a FREEFORM tool, so do not wrap the pat...; hash=dea9f12b0111a4ba
- function/view_image: View a local image file from the filesystem when visual inspection is needed. Use this for images...; args=[path]; hash=1908de9c0be7118a
- namespace/multi_agent_v1: Tools for spawning and managing sub-agents.; hash=cb54fdcce4eb403b
- namespace/multi_agent_v1: Tools for spawning and managing sub-agents.; hash=616c9e490b1773bd
- function/close_agent
- function/resume_agent
- function/send_input
@@ -15,7 +15,7 @@ Settings:
stream: true
text: {"verbosity":"low"}
tool_choice: "auto"
tools (12; hash=249f1e48c93d8aaf):
tools (12; hash=f5a7b6121039ed67):
- function/exec_command: Runs a command in a PTY, returning output or a session ID for ongoing interaction.; args=[cmd, justification, login, max_output_tokens, prefix_rule, sandbox_permissions, shell, tty, workdir, yield_time_ms]; hash=24f4b6739dd5d940
- function/write_stdin: Writes characters to an existing unified exec session and returns recent output.; args=[chars, max_output_tokens, session_id, yield_time_ms]; hash=7fb62d8a3e2df4e2
- function/list_mcp_resources: Lists resources provided by MCP servers. Resources allow servers to share data that provides cont...; args=[cursor, server]; hash=6792edaff13561bd
@@ -24,7 +24,7 @@ Settings:
- function/request_user_input: Request user input for one to three short questions and wait for the response. This tool is only ...; args=[questions]; hash=64a729b25c424fc6
- custom/apply_patch: The `apply_patch` tool can be used to edit files. This is a FREEFORM tool, so do not wrap the pat...; hash=dea9f12b0111a4ba
- function/view_image: View a local image file from the filesystem when visual inspection is needed. Use this for images...; args=[path]; hash=1908de9c0be7118a
- namespace/multi_agent_v1: Tools for spawning and managing sub-agents.; hash=cb54fdcce4eb403b
- namespace/multi_agent_v1: Tools for spawning and managing sub-agents.; hash=616c9e490b1773bd
- function/close_agent
- function/resume_agent
- function/send_input
@@ -15,7 +15,7 @@ Settings:
stream: true
text: {"verbosity":"low"}
tool_choice: "auto"
tools (12; hash=249f1e48c93d8aaf):
tools (12; hash=f5a7b6121039ed67):
- function/exec_command: Runs a command in a PTY, returning output or a session ID for ongoing interaction.; args=[cmd, justification, login, max_output_tokens, prefix_rule, sandbox_permissions, shell, tty, workdir, yield_time_ms]; hash=24f4b6739dd5d940
- function/write_stdin: Writes characters to an existing unified exec session and returns recent output.; args=[chars, max_output_tokens, session_id, yield_time_ms]; hash=7fb62d8a3e2df4e2
- function/list_mcp_resources: Lists resources provided by MCP servers. Resources allow servers to share data that provides cont...; args=[cursor, server]; hash=6792edaff13561bd
@@ -24,7 +24,7 @@ Settings:
- function/request_user_input: Request user input for one to three short questions and wait for the response. This tool is only ...; args=[questions]; hash=64a729b25c424fc6
- custom/apply_patch: The `apply_patch` tool can be used to edit files. This is a FREEFORM tool, so do not wrap the pat...; hash=dea9f12b0111a4ba
- function/view_image: View a local image file from the filesystem when visual inspection is needed. Use this for images...; args=[path]; hash=1908de9c0be7118a
- namespace/multi_agent_v1: Tools for spawning and managing sub-agents.; hash=cb54fdcce4eb403b
- namespace/multi_agent_v1: Tools for spawning and managing sub-agents.; hash=616c9e490b1773bd
- function/close_agent
- function/resume_agent
- function/send_input
@@ -16,7 +16,7 @@ Settings:
text: {"verbosity":"low"}
tool_choice: "auto"
-- request 1 (turn) --
00:additional_tools/developer (3; hash=e65af59b227d000c):
00:additional_tools/developer (3; hash=2cefe21beec1e2d2):
- namespace/functions; hash=1377b145cdccf5b4
- custom/exec
- function/wait
@@ -24,7 +24,7 @@ Settings:
- function/request_user_input_async
- namespace/clock: Tools for reading and waiting on time.; hash=f402e8c5e9b5e317
- function/sleep
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=0271b89d55152d99
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=a097b9364047b8bb
- function/followup_task
- function/interrupt_agent
- function/list_agents
@@ -54,14 +54,14 @@ Settings:
## Window 2 (after request 3: input diverged at item 00)
Settings: same as window 1
-- request 4 (turn) --
00:additional_tools/developer (3; hash=4c3fa46ab2c6ed7d):
00:additional_tools/developer (3; hash=72728edaee929e95):
- namespace/functions; hash=d0883eefc88509b7
- custom/exec
- function/wait
- function/request_user_input
- namespace/clock: Tools for reading and waiting on time.; hash=f402e8c5e9b5e317
- function/sleep
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=0271b89d55152d99
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=a097b9364047b8bb
- function/followup_task
- function/interrupt_agent
- function/list_agents
@@ -15,14 +15,14 @@ Settings:
stream: true
text: {"verbosity":"low"}
tool_choice: "auto"
tools (7; hash=599ab6aae9a806fb):
tools (7; hash=4d27bfb639149cfb):
- custom/exec: Catalog exec instructions. Default yield: 1234 ms. `image(imageUrlOrItem: string | { image_url: s...; hash=26fc765b2c6fa4eb
- function/wait: Catalog wait instructions: resume the yielded cell.; args=[cell_id, max_tokens, terminate, yield_time_ms]; hash=67bc0f04da3505b3
- function/request_user_input: Request user input for one to three short questions and wait for the response. This tool is only ...; args=[questions]; hash=64a729b25c424fc6
- function/request_user_input_async: Ask the user one or more questions during ongoing work. Use this tool only to request missing inf...; args=[questions]; hash=d6d74f33db330162
- namespace/clock: Tools for reading and waiting on time.; hash=f402e8c5e9b5e317
- function/sleep
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=0271b89d55152d99
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=a097b9364047b8bb
- function/followup_task
- function/interrupt_agent
- function/list_agents
@@ -40,14 +40,14 @@ Settings:
User authorization and preferences persist across turns. Do not request permission again when the user h... any guidelines provided in skills or external files. [hash=911f91c92c72882c]
<OMITTED 155 LINES; ~4935 TOKENS; hash=5573869716192298>
<OMITTED 151 LINES; ~4851 TOKENS; hash=959762226fedfd0c>
## Decision Boundary
- Skill naming: If a plugin contributes skills, those skill entries are prefixed with plugin_name: in the Skills list.
- MCP naming: Plugin-provided MCP tools keep standard MCP identifiers such as mcp__server__tool; use tool provenance to tell which plugin they come from.
- Trigger rules: If the user explicitly names a plugin, prefer capabilities associated with that plugin for that turn.
- Relationship to capabilities: Plugins are not invoked directly. Use their underlying skills, MCP tools, and app tools to help solve the task.
- Relevance: Determine what a plugin can help with from explicit user mention or from the plugin-associa..., MCP tools, and apps exposed elsewhere in this turn. [hash=620459a4e7b590f0]
- Missing/blocked: If the user requests a plugin that does not have relevant callable capabilities for t..., say so briefly and continue with the best fallback. [hash=746b0469a8f6cb3e]
Discover and use relevant tools proactively when they may help carry out the requested task. For example...ng the user for information they may already contain. [hash=3af5c93cbc00b4c7]
When a plugin is available for a service and its tools are sufficient for completing the task, prefer th...er has an explicit preference, defer to that instead. [hash=936dbd7fe76c210f]
A user can also tag a plugin using `[@plugin-name](plugin://{{plugin_id}})` or an app using `[$app-name]... app, prefer its relevant capabilities for that turn. [hash=8730f69f41534f8f]
01:message/developer[3]:
[01] <permissions instructions>
@@ -15,14 +15,14 @@ Settings:
stream: true
text: {"verbosity":"low"}
tool_choice: "auto"
tools (7; hash=2520831eb71eb406):
tools (7; hash=a2dd9deb36b0b574):
- custom/exec: Catalog exec instructions. Default yield: 1234 ms. `image(imageUrlOrItem: string | { image_url: s...; hash=043d7068a7445992
- function/wait: Catalog wait instructions: resume the yielded cell.; args=[cell_id, max_tokens, terminate, yield_time_ms]; hash=67bc0f04da3505b3
- function/request_user_input: Request user input for one to three short questions and wait for the response. This tool is only ...; args=[questions]; hash=64a729b25c424fc6
- function/request_user_input_async: Ask the user one or more questions during ongoing work. Use this tool only to request missing inf...; args=[questions]; hash=d6d74f33db330162
- namespace/clock: Tools for reading and waiting on time.; hash=f402e8c5e9b5e317
- function/sleep
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=0271b89d55152d99
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=a097b9364047b8bb
- function/followup_task
- function/interrupt_agent
- function/list_agents
@@ -40,14 +40,14 @@ Settings:
User authorization and preferences persist across turns. Do not request permission again when the user h... any guidelines provided in skills or external files. [hash=911f91c92c72882c]
<OMITTED 155 LINES; ~4935 TOKENS; hash=5573869716192298>
<OMITTED 151 LINES; ~4851 TOKENS; hash=959762226fedfd0c>
## Decision Boundary
- Skill naming: If a plugin contributes skills, those skill entries are prefixed with plugin_name: in the Skills list.
- MCP naming: Plugin-provided MCP tools keep standard MCP identifiers such as mcp__server__tool; use tool provenance to tell which plugin they come from.
- Trigger rules: If the user explicitly names a plugin, prefer capabilities associated with that plugin for that turn.
- Relationship to capabilities: Plugins are not invoked directly. Use their underlying skills, MCP tools, and app tools to help solve the task.
- Relevance: Determine what a plugin can help with from explicit user mention or from the plugin-associa..., MCP tools, and apps exposed elsewhere in this turn. [hash=620459a4e7b590f0]
- Missing/blocked: If the user requests a plugin that does not have relevant callable capabilities for t..., say so briefly and continue with the best fallback. [hash=746b0469a8f6cb3e]
Discover and use relevant tools proactively when they may help carry out the requested task. For example...ng the user for information they may already contain. [hash=3af5c93cbc00b4c7]
When a plugin is available for a service and its tools are sufficient for completing the task, prefer th...er has an explicit preference, defer to that instead. [hash=936dbd7fe76c210f]
A user can also tag a plugin using `[@plugin-name](plugin://{{plugin_id}})` or an app using `[$app-name]... app, prefer its relevant capabilities for that turn. [hash=8730f69f41534f8f]
01:message/developer[3]:
[01] <permissions instructions>
@@ -15,8 +15,8 @@ Settings:
stream: true
text: {"verbosity":"low"}
tool_choice: "auto"
tools (4; hash=13b656e85aa8469e):
- custom/exec: Run JavaScript code to orchestrate/compose tool calls - Evaluates the provided JavaScript code in...; hash=f1d1e5487f896635
tools (4; hash=6f9152dcf0de454e):
- custom/exec: Run JavaScript code to orchestrate/compose tool calls - Evaluates the provided JavaScript code in...; hash=386d93d73548c42f
- function/wait: Waits on a yielded `exec` cell and returns new output or completion. - Use `wait` only after `exe...; args=[cell_id, max_tokens, terminate, yield_time_ms]; hash=6692d8f6c55be489
- function/request_user_input: Request user input for one to three short questions and wait for the response. This tool is only ...; args=[questions]; hash=64a729b25c424fc6
- web_search; hash=e214d4db1bc49e89
@@ -6,7 +6,7 @@ Scenario: Settlement helpers deliver fulfilled and rejected records with input i
## Window 1
-- request 1 (turn) --
00:additional_tools/developer (3; hash=f2e4edd02b6f9300):
00:additional_tools/developer (3; hash=a7c46a0959d54866):
- namespace/functions; hash=81846de5bc4e7474
- custom/exec
- function/wait
@@ -14,7 +14,7 @@ Scenario: Settlement helpers deliver fulfilled and rejected records with input i
- function/request_user_input_async
- namespace/clock: Tools for reading and waiting on time.; hash=f402e8c5e9b5e317
- function/sleep
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=0271b89d55152d99
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=a097b9364047b8bb
- function/followup_task
- function/interrupt_agent
- function/list_agents
@@ -6,7 +6,7 @@ Scenario: A Code Mode script drops its command handle. WorldState retains it whi
## Window 1
-- request 1 (turn) --
00:additional_tools/developer (3; hash=a4fdf355eb986849):
00:additional_tools/developer (3; hash=08fdf601acba9636):
- namespace/functions; hash=81846de5bc4e7474
- custom/exec
- function/wait
@@ -14,7 +14,7 @@ Scenario: A Code Mode script drops its command handle. WorldState retains it whi
- function/request_user_input_async
- namespace/clock: Tools for reading and waiting on time.; hash=f402e8c5e9b5e317
- function/sleep
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=299e38db0471a16a
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=132d0256c7f80203
- function/followup_task
- function/interrupt_agent
- function/list_agents
@@ -58,7 +58,7 @@ Scenario: A Code Mode script drops its command handle. WorldState retains it whi
## Window 2 (after request 3: settings changed (model))
-- request 4 (turn) --
00:additional_tools/developer (3; hash=a4fdf355eb986849):
00:additional_tools/developer (3; hash=08fdf601acba9636):
- namespace/functions; hash=81846de5bc4e7474
- custom/exec
- function/wait
@@ -66,7 +66,7 @@ Scenario: A Code Mode script drops its command handle. WorldState retains it whi
- function/request_user_input_async
- namespace/clock: Tools for reading and waiting on time.; hash=f402e8c5e9b5e317
- function/sleep
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=299e38db0471a16a
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=132d0256c7f80203
- function/followup_task
- function/interrupt_agent
- function/list_agents
@@ -113,7 +113,7 @@ Scenario: A Code Mode script drops its command handle. WorldState retains it whi
</command_sessions>
13:message/developer:
This is an incremental tools update. Previously declared tools remain available for direct calls unless ...here, its latest definition replaces the earlier one. [hash=bceb65fad1afdf9c]
14:additional_tools/developer (1; hash=2a7a524a5c95d741):
14:additional_tools/developer (1; hash=5313c72115e5e96e):
- namespace/functions; hash=429e5891e0f9c86b
- custom/exec
15:message/developer:
@@ -130,7 +130,7 @@ Scenario: A Code Mode script drops its command handle. WorldState retains it whi
## Window 3 (after request 5: input diverged at item 00, settings changed (model))
-- request 6 (turn) --
00:additional_tools/developer (3; hash=190749c3ef070118):
00:additional_tools/developer (3; hash=d5796ee97b1e1527):
- namespace/functions; hash=f6f779209c4af0f3
- custom/exec
- function/wait
@@ -138,7 +138,7 @@ Scenario: A Code Mode script drops its command handle. WorldState retains it whi
- function/request_user_input_async
- namespace/clock: Tools for reading and waiting on time.; hash=f402e8c5e9b5e317
- function/sleep
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=299e38db0471a16a
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=132d0256c7f80203
- function/followup_task
- function/interrupt_agent
- function/list_agents
@@ -175,7 +175,7 @@ Scenario: A Code Mode script drops its command handle. WorldState retains it whi
</command_sessions>
11:message/developer:
This is an incremental tools update. Previously declared tools remain available for direct calls unless ...here, its latest definition replaces the earlier one. [hash=bceb65fad1afdf9c]
12:additional_tools/developer (1; hash=fd836d9e37127de2):
12:additional_tools/developer (1; hash=b762b19e2542cbbd):
- namespace/functions; hash=04341bdad1d8e6f6
- custom/exec
13:message/developer:
@@ -16,7 +16,7 @@ Settings:
text: {"verbosity":"low"}
tool_choice: "auto"
-- request 1 (turn; Before /compact) --
00:additional_tools/developer (3; hash=f2e4edd02b6f9300):
00:additional_tools/developer (3; hash=a7c46a0959d54866):
- namespace/functions; hash=81846de5bc4e7474
- custom/exec
- function/wait
@@ -24,7 +24,7 @@ Settings:
- function/request_user_input_async
- namespace/clock: Tools for reading and waiting on time.; hash=f402e8c5e9b5e317
- function/sleep
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=0271b89d55152d99
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=a097b9364047b8bb
- function/followup_task
- function/interrupt_agent
- function/list_agents
@@ -40,14 +40,14 @@ Settings:
User authorization and preferences persist across turns. Do not request permission again when the user h... any guidelines provided in skills or external files. [hash=911f91c92c72882c]
<OMITTED 155 LINES; ~4935 TOKENS; hash=5573869716192298>
<OMITTED 151 LINES; ~4851 TOKENS; hash=959762226fedfd0c>
## Decision Boundary
- Skill naming: If a plugin contributes skills, those skill entries are prefixed with plugin_name: in the Skills list.
- MCP naming: Plugin-provided MCP tools keep standard MCP identifiers such as mcp__server__tool; use tool provenance to tell which plugin they come from.
- Trigger rules: If the user explicitly names a plugin, prefer capabilities associated with that plugin for that turn.
- Relationship to capabilities: Plugins are not invoked directly. Use their underlying skills, MCP tools, and app tools to help solve the task.
- Relevance: Determine what a plugin can help with from explicit user mention or from the plugin-associa..., MCP tools, and apps exposed elsewhere in this turn. [hash=620459a4e7b590f0]
- Missing/blocked: If the user requests a plugin that does not have relevant callable capabilities for t..., say so briefly and continue with the best fallback. [hash=746b0469a8f6cb3e]
Discover and use relevant tools proactively when they may help carry out the requested task. For example...ng the user for information they may already contain. [hash=3af5c93cbc00b4c7]
When a plugin is available for a service and its tools are sufficient for completing the task, prefer th...er has an explicit preference, defer to that instead. [hash=936dbd7fe76c210f]
A user can also tag a plugin using `[@plugin-name](plugin://{{plugin_id}})` or an app using `[$app-name]... app, prefer its relevant capabilities for that turn. [hash=8730f69f41534f8f]
02:message/developer[2]:
[01] <permissions instructions>
@@ -112,7 +112,7 @@ Settings:
## Window 2 (after request 2: input diverged at item 07)
Settings: same as window 1
-- request 3 (compaction; Repeated /compact after resume) --
00:additional_tools/developer (3; hash=f2e4edd02b6f9300):
00:additional_tools/developer (3; hash=a7c46a0959d54866):
- namespace/functions; hash=81846de5bc4e7474
- custom/exec
- function/wait
@@ -120,7 +120,7 @@ Settings: same as window 1
- function/request_user_input_async
- namespace/clock: Tools for reading and waiting on time.; hash=f402e8c5e9b5e317
- function/sleep
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=0271b89d55152d99
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=a097b9364047b8bb
- function/followup_task
- function/interrupt_agent
- function/list_agents
@@ -136,14 +136,14 @@ Settings: same as window 1
User authorization and preferences persist across turns. Do not request permission again when the user h... any guidelines provided in skills or external files. [hash=911f91c92c72882c]
<OMITTED 155 LINES; ~4935 TOKENS; hash=5573869716192298>
<OMITTED 151 LINES; ~4851 TOKENS; hash=959762226fedfd0c>
## Decision Boundary
- Skill naming: If a plugin contributes skills, those skill entries are prefixed with plugin_name: in the Skills list.
- MCP naming: Plugin-provided MCP tools keep standard MCP identifiers such as mcp__server__tool; use tool provenance to tell which plugin they come from.
- Trigger rules: If the user explicitly names a plugin, prefer capabilities associated with that plugin for that turn.
- Relationship to capabilities: Plugins are not invoked directly. Use their underlying skills, MCP tools, and app tools to help solve the task.
- Relevance: Determine what a plugin can help with from explicit user mention or from the plugin-associa..., MCP tools, and apps exposed elsewhere in this turn. [hash=620459a4e7b590f0]
- Missing/blocked: If the user requests a plugin that does not have relevant callable capabilities for t..., say so briefly and continue with the best fallback. [hash=746b0469a8f6cb3e]
Discover and use relevant tools proactively when they may help carry out the requested task. For example...ng the user for information they may already contain. [hash=3af5c93cbc00b4c7]
When a plugin is available for a service and its tools are sufficient for completing the task, prefer th...er has an explicit preference, defer to that instead. [hash=936dbd7fe76c210f]
A user can also tag a plugin using `[@plugin-name](plugin://{{plugin_id}})` or an app using `[$app-name]... app, prefer its relevant capabilities for that turn. [hash=8730f69f41534f8f]
02:message/developer[2]:
[01] <permissions instructions>
@@ -205,7 +205,7 @@ Settings: same as window 1
## Window 3 (after request 3: input diverged at item 02)
Settings: same as window 1
-- request 4 (turn; First turn after /compact) --
00:additional_tools/developer (3; hash=f2e4edd02b6f9300):
00:additional_tools/developer (3; hash=a7c46a0959d54866):
- namespace/functions; hash=81846de5bc4e7474
- custom/exec
- function/wait
@@ -213,7 +213,7 @@ Settings: same as window 1
- function/request_user_input_async
- namespace/clock: Tools for reading and waiting on time.; hash=f402e8c5e9b5e317
- function/sleep
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=0271b89d55152d99
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=a097b9364047b8bb
- function/followup_task
- function/interrupt_agent
- function/list_agents
@@ -229,14 +229,14 @@ Settings: same as window 1
User authorization and preferences persist across turns. Do not request permission again when the user h... any guidelines provided in skills or external files. [hash=911f91c92c72882c]
<OMITTED 155 LINES; ~4935 TOKENS; hash=5573869716192298>
<OMITTED 151 LINES; ~4851 TOKENS; hash=959762226fedfd0c>
## Decision Boundary
- Skill naming: If a plugin contributes skills, those skill entries are prefixed with plugin_name: in the Skills list.
- MCP naming: Plugin-provided MCP tools keep standard MCP identifiers such as mcp__server__tool; use tool provenance to tell which plugin they come from.
- Trigger rules: If the user explicitly names a plugin, prefer capabilities associated with that plugin for that turn.
- Relationship to capabilities: Plugins are not invoked directly. Use their underlying skills, MCP tools, and app tools to help solve the task.
- Relevance: Determine what a plugin can help with from explicit user mention or from the plugin-associa..., MCP tools, and apps exposed elsewhere in this turn. [hash=620459a4e7b590f0]
- Missing/blocked: If the user requests a plugin that does not have relevant callable capabilities for t..., say so briefly and continue with the best fallback. [hash=746b0469a8f6cb3e]
Discover and use relevant tools proactively when they may help carry out the requested task. For example...ng the user for information they may already contain. [hash=3af5c93cbc00b4c7]
When a plugin is available for a service and its tools are sufficient for completing the task, prefer th...er has an explicit preference, defer to that instead. [hash=936dbd7fe76c210f]
A user can also tag a plugin using `[@plugin-name](plugin://{{plugin_id}})` or an app using `[$app-name]... app, prefer its relevant capabilities for that turn. [hash=8730f69f41534f8f]
02:message/developer[2]:
[01] <permissions instructions>
@@ -1,6 +1,5 @@
---
source: core/tests/suite/scenarios_dynamic_tools.rs
assertion_line: 173
expression: "context_snapshot::format_request_history_snapshot(\"A pending release call survives replacement with deployment. The next step uses deployment. Replacement and clearing each survive restart.\",\n&requests,\n&ContextSnapshotOptions::default().rewrite_known_segments().include_request_settings(),)"
---
Scenario: A pending release call survives replacement with deployment. The next step uses deployment. Replacement and clearing each survive restart.
@@ -16,7 +15,7 @@ Settings:
stream: true
text: {"verbosity":"low"}
tool_choice: "auto"
tools (10; hash=e91eda48fd5bd29e):
tools (10; hash=926cde2726c284b6):
- function/exec_command: Runs a command in a PTY, returning output or a session ID for ongoing interaction.; args=[cmd, justification, login, max_output_tokens, prefix_rule, sandbox_permissions, shell, tty, workdir, yield_time_ms]; hash=24f4b6739dd5d940
- function/write_stdin: Writes characters to an existing unified exec session and returns recent output.; args=[chars, max_output_tokens, session_id, yield_time_ms]; hash=7fb62d8a3e2df4e2
- function/request_user_input: Request user input for one to three short questions and wait for the response. This tool is only ...; args=[questions]; hash=64a729b25c424fc6
@@ -26,7 +25,7 @@ Settings:
- function/sleep
- custom/apply_patch: The `apply_patch` tool can be used to edit files. This is a FREEFORM tool, so do not wrap the pat...; hash=dea9f12b0111a4ba
- function/view_image: View a local image file from the filesystem when visual inspection is needed. Use this for images...; args=[path]; hash=1908de9c0be7118a
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=0271b89d55152d99
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=a097b9364047b8bb
- function/followup_task
- function/interrupt_agent
- function/list_agents
@@ -16,7 +16,7 @@ Settings:
text: {"verbosity":"low"}
tool_choice: "auto"
-- request 1 (request) --
00:additional_tools/developer (3; hash=e37d13847f694d9d):
00:additional_tools/developer (3; hash=8f90f1c7f4fce935):
- namespace/functions; hash=3eb85b1b397a4861
- custom/exec
- function/wait
@@ -24,7 +24,7 @@ Settings:
- function/request_user_input_async
- namespace/clock: Tools for reading and waiting on time.; hash=f402e8c5e9b5e317
- function/sleep
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=0271b89d55152d99
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=a097b9364047b8bb
- function/followup_task
- function/interrupt_agent
- function/list_agents
@@ -40,14 +40,14 @@ Settings:
User authorization and preferences persist across turns. Do not request permission again when the user h... any guidelines provided in skills or external files. [hash=911f91c92c72882c]
<OMITTED 155 LINES; ~4935 TOKENS; hash=5573869716192298>
<OMITTED 151 LINES; ~4851 TOKENS; hash=959762226fedfd0c>
## Decision Boundary
- Skill naming: If a plugin contributes skills, those skill entries are prefixed with plugin_name: in the Skills list.
- MCP naming: Plugin-provided MCP tools keep standard MCP identifiers such as mcp__server__tool; use tool provenance to tell which plugin they come from.
- Trigger rules: If the user explicitly names a plugin, prefer capabilities associated with that plugin for that turn.
- Relationship to capabilities: Plugins are not invoked directly. Use their underlying skills, MCP tools, and app tools to help solve the task.
- Relevance: Determine what a plugin can help with from explicit user mention or from the plugin-associa..., MCP tools, and apps exposed elsewhere in this turn. [hash=620459a4e7b590f0]
- Missing/blocked: If the user requests a plugin that does not have relevant callable capabilities for t..., say so briefly and continue with the best fallback. [hash=746b0469a8f6cb3e]
Discover and use relevant tools proactively when they may help carry out the requested task. For example...ng the user for information they may already contain. [hash=3af5c93cbc00b4c7]
When a plugin is available for a service and its tools are sufficient for completing the task, prefer th...er has an explicit preference, defer to that instead. [hash=936dbd7fe76c210f]
A user can also tag a plugin using `[@plugin-name](plugin://{{plugin_id}})` or an app using `[$app-name]... app, prefer its relevant capabilities for that turn. [hash=8730f69f41534f8f]
02:message/developer[3]:
[01] <permissions instructions>
@@ -1,6 +1,6 @@
---
source: core/tests/suite/scenarios_incremental_tools.rs
expression: "context_snapshot::format_request_history_snapshot(\"MCP refreshes remove one tool, then its whole namespace, before restoring both tools. Removal notices append alone; restoration appends one developer notice before both namespace declarations. Unchanged turns do not repeat updates.\",\n&requests,\n&ContextSnapshotOptions::default().rewrite_known_segments().include_request_settings(),)"
expression: "context_snapshot::format_request_history_snapshot(scenario, &requests,\n&ContextSnapshotOptions::default().rewrite_known_segments().include_request_settings(),)"
---
Scenario: MCP refreshes remove one tool, then its whole namespace, before restoring both tools. Removal notices append alone; restoration appends one developer notice before both namespace declarations. Unchanged turns do not repeat updates.
@@ -16,7 +16,7 @@ Settings:
text: {"verbosity":"low"}
tool_choice: "auto"
-- request 1 (turn) --
00:additional_tools/developer (3; hash=62bd6d863208b8d8):
00:additional_tools/developer (3; hash=9af04ba87a0a3228):
- namespace/functions; hash=8f4955ac0f6263ad
- function/exec_command
- function/write_stdin
@@ -26,7 +26,7 @@ Settings:
- function/request_user_input
- custom/apply_patch
- function/view_image
- namespace/multi_agent_v1: Tools for spawning and managing sub-agents.; hash=f066a006a1fd060e
- namespace/multi_agent_v1: Tools for spawning and managing sub-agents.; hash=c041c3bd4f63167c
- function/close_agent
- function/resume_agent
- function/send_input
@@ -1,6 +1,5 @@
---
source: core/tests/suite/scenarios_incremental_tools.rs
assertion_line: 644
expression: "context_snapshot::format_request_history_snapshot(scenario, &requests,\n&ContextSnapshotOptions::default().rewrite_known_segments().include_request_settings(),)"
---
Scenario: MCP refreshes remove one tool, then its whole namespace, before restoring both tools. Removal notices append alone; restoration appends one developer notice before both namespace declarations. Unchanged turns do not repeat updates.
@@ -19,7 +18,7 @@ Settings:
tools (1; hash=f359ca6f02a419d6):
- web_search; hash=f103e327c6870a46
-- request 1 (turn) --
00:additional_tools/developer (3; hash=62bd6d863208b8d8):
00:additional_tools/developer (3; hash=9af04ba87a0a3228):
- namespace/functions; hash=8f4955ac0f6263ad
- function/exec_command
- function/write_stdin
@@ -29,7 +28,7 @@ Settings:
- function/request_user_input
- custom/apply_patch
- function/view_image
- namespace/multi_agent_v1: Tools for spawning and managing sub-agents.; hash=f066a006a1fd060e
- namespace/multi_agent_v1: Tools for spawning and managing sub-agents.; hash=c041c3bd4f63167c
- function/close_agent
- function/resume_agent
- function/send_input
@@ -16,7 +16,7 @@ Settings:
text: {"verbosity":"low"}
tool_choice: "auto"
-- request 1 (turn) --
00:additional_tools/developer (3; hash=62bd6d863208b8d8):
00:additional_tools/developer (3; hash=9af04ba87a0a3228):
- namespace/functions; hash=8f4955ac0f6263ad
- function/exec_command
- function/write_stdin
@@ -26,7 +26,7 @@ Settings:
- function/request_user_input
- custom/apply_patch
- function/view_image
- namespace/multi_agent_v1: Tools for spawning and managing sub-agents.; hash=f066a006a1fd060e
- namespace/multi_agent_v1: Tools for spawning and managing sub-agents.; hash=c041c3bd4f63167c
- function/close_agent
- function/resume_agent
- function/send_input
@@ -15,7 +15,7 @@ Settings:
stream: true
text: {"verbosity":"low"}
tool_choice: "auto"
tools (13; hash=2469039dcc5dff80):
tools (13; hash=ec9204e933ea40da):
- function/exec_command: Runs a command in a PTY, returning output or a session ID for ongoing interaction.; args=[cmd, justification, login, max_output_tokens, prefix_rule, sandbox_permissions, shell, tty, workdir, yield_time_ms]; hash=24f4b6739dd5d940
- function/write_stdin: Writes characters to an existing unified exec session and returns recent output.; args=[chars, max_output_tokens, session_id, yield_time_ms]; hash=7fb62d8a3e2df4e2
- function/list_mcp_resources: Lists resources provided by MCP servers. Resources allow servers to share data that provides cont...; args=[cursor, server]; hash=6792edaff13561bd
@@ -28,7 +28,7 @@ Settings:
- function/sleep
- custom/apply_patch: The `apply_patch` tool can be used to edit files. This is a FREEFORM tool, so do not wrap the pat...; hash=dea9f12b0111a4ba
- function/view_image: View a local image file from the filesystem when visual inspection is needed. Use this for images...; args=[path]; hash=1908de9c0be7118a
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=0271b89d55152d99
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=a097b9364047b8bb
- function/followup_task
- function/interrupt_agent
- function/list_agents
@@ -15,7 +15,7 @@ Settings:
stream: true
text: {"verbosity":"low"}
tool_choice: "auto"
tools (14; hash=6009d2b483bffccf):
tools (14; hash=07d10df366aa7fc5):
- function/exec_command: Runs a command in a PTY, returning output or a session ID for ongoing interaction.; args=[cmd, justification, login, max_output_tokens, prefix_rule, sandbox_permissions, shell, tty, workdir, yield_time_ms]; hash=24f4b6739dd5d940
- function/write_stdin: Writes characters to an existing unified exec session and returns recent output.; args=[chars, max_output_tokens, session_id, yield_time_ms]; hash=7fb62d8a3e2df4e2
- function/list_mcp_resources: Lists resources provided by MCP servers. Resources allow servers to share data that provides cont...; args=[cursor, server]; hash=6792edaff13561bd
@@ -28,7 +28,7 @@ Settings:
- function/sleep
- custom/apply_patch: The `apply_patch` tool can be used to edit files. This is a FREEFORM tool, so do not wrap the pat...; hash=dea9f12b0111a4ba
- function/view_image: View a local image file from the filesystem when visual inspection is needed. Use this for images...; args=[path]; hash=1908de9c0be7118a
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=0271b89d55152d99
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=a097b9364047b8bb
- function/followup_task
- function/interrupt_agent
- function/list_agents
@@ -67,7 +67,7 @@ Settings:
## Window 2 (after request 2)
Settings: relative to window 1
tools (13; hash=2469039dcc5dff80):
tools (13; hash=ec9204e933ea40da):
- removed namespace/mcp__enterprise
-- request 3 (turn) --
00:message/developer:
@@ -15,13 +15,13 @@ Settings:
stream: true
text: {"verbosity":"low"}
tool_choice: "auto"
tools (6; hash=fd811654963b2f85):
tools (6; hash=d843621dc36edf6d):
- custom/exec: Run JavaScript code to orchestrate/compose tool calls - Evaluates the provided JavaScript code in...; hash=c00e260794631906
- function/wait: Waits on a yielded `exec` cell and returns new output or completion. - Use `wait` only after `exe...; args=[cell_id, max_tokens, terminate, yield_time_ms]; hash=6692d8f6c55be489
- function/request_user_input: Request user input for one to three short questions and wait for the response. This tool is only ...; args=[questions]; hash=64a729b25c424fc6
- namespace/clock: Tools for reading and waiting on time.; hash=f402e8c5e9b5e317
- function/sleep
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=0271b89d55152d99
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=a097b9364047b8bb
- function/followup_task
- function/interrupt_agent
- function/list_agents
@@ -39,14 +39,14 @@ Settings:
User authorization and preferences persist across turns. Do not request permission again when the user h... any guidelines provided in skills or external files. [hash=911f91c92c72882c]
<OMITTED 155 LINES; ~4935 TOKENS; hash=5573869716192298>
<OMITTED 151 LINES; ~4851 TOKENS; hash=959762226fedfd0c>
## Decision Boundary
- Skill naming: If a plugin contributes skills, those skill entries are prefixed with plugin_name: in the Skills list.
- MCP naming: Plugin-provided MCP tools keep standard MCP identifiers such as mcp__server__tool; use tool provenance to tell which plugin they come from.
- Trigger rules: If the user explicitly names a plugin, prefer capabilities associated with that plugin for that turn.
- Relationship to capabilities: Plugins are not invoked directly. Use their underlying skills, MCP tools, and app tools to help solve the task.
- Relevance: Determine what a plugin can help with from explicit user mention or from the plugin-associa..., MCP tools, and apps exposed elsewhere in this turn. [hash=620459a4e7b590f0]
- Missing/blocked: If the user requests a plugin that does not have relevant callable capabilities for t..., say so briefly and continue with the best fallback. [hash=746b0469a8f6cb3e]
Discover and use relevant tools proactively when they may help carry out the requested task. For example...ng the user for information they may already contain. [hash=3af5c93cbc00b4c7]
When a plugin is available for a service and its tools are sufficient for completing the task, prefer th...er has an explicit preference, defer to that instead. [hash=936dbd7fe76c210f]
A user can also tag a plugin using `[@plugin-name](plugin://{{plugin_id}})` or an app using `[$app-name]... app, prefer its relevant capabilities for that turn. [hash=8730f69f41534f8f]
01:message/developer[3]:
[01] <permissions instructions>
@@ -1,6 +1,6 @@
---
source: core/tests/suite/scenarios_mcp_sharing.rs
expression: "context_snapshot::format_context_snapshot(\"Two chats share one MCP process, expose different tools, and answer simultaneous model-requested questions in reverse order.\",\n&entries,\n&ContextSnapshotOptions::default().rewrite_known_segments().include_request_settings(),)"
expression: snapshot
---
Scenario: Two chats share one MCP process, expose different tools, and answer simultaneous model-requested questions in reverse order.
@@ -15,7 +15,7 @@ Settings:
stream: true
text: {"verbosity":"low"}
tool_choice: "auto"
tools (11; hash=41fdc4933aaabae7):
tools (11; hash=d5473b1cfe0a4e6f):
- function/exec_command: Runs a command in a PTY, returning output or a session ID for ongoing interaction.; args=[cmd, justification, login, max_output_tokens, prefix_rule, sandbox_permissions, shell, tty, workdir, yield_time_ms]; hash=24f4b6739dd5d940
- function/write_stdin: Writes characters to an existing unified exec session and returns recent output.; args=[chars, max_output_tokens, session_id, yield_time_ms]; hash=7fb62d8a3e2df4e2
- function/list_mcp_resources: Lists resources provided by MCP servers. Resources allow servers to share data that provides cont...; args=[cursor, server]; hash=6792edaff13561bd
@@ -24,7 +24,7 @@ Settings:
- function/request_user_input: Request user input for one to three short questions and wait for the response. This tool is only ...; args=[questions]; hash=64a729b25c424fc6
- custom/apply_patch: The `apply_patch` tool can be used to edit files. This is a FREEFORM tool, so do not wrap the pat...; hash=dea9f12b0111a4ba
- function/view_image: View a local image file from the filesystem when visual inspection is needed. Use this for images...; args=[path]; hash=1908de9c0be7118a
- namespace/multi_agent_v1: Tools for spawning and managing sub-agents.; hash=f066a006a1fd060e
- namespace/multi_agent_v1: Tools for spawning and managing sub-agents.; hash=c041c3bd4f63167c
- function/close_agent
- function/resume_agent
- function/send_input
@@ -15,7 +15,7 @@ Settings:
stream: true
text: {"verbosity":"low"}
tool_choice: "auto"
tools (11; hash=8ba891fa89d0963e):
tools (11; hash=9672d359145082c6):
- function/exec_command: Runs a command in a PTY, returning output or a session ID for ongoing interaction.; args=[cmd, justification, login, max_output_tokens, prefix_rule, sandbox_permissions, shell, tty, workdir, yield_time_ms]; hash=24f4b6739dd5d940
- function/write_stdin: Writes characters to an existing unified exec session and returns recent output.; args=[chars, max_output_tokens, session_id, yield_time_ms]; hash=7fb62d8a3e2df4e2
- function/list_mcp_resources: Lists resources provided by MCP servers. Resources allow servers to share data that provides cont...; args=[cursor, server]; hash=6792edaff13561bd
@@ -24,7 +24,7 @@ Settings:
- function/request_user_input: Request user input for one to three short questions and wait for the response. This tool is only ...; args=[questions]; hash=64a729b25c424fc6
- custom/apply_patch: The `apply_patch` tool can be used to edit files. This is a FREEFORM tool, so do not wrap the pat...; hash=dea9f12b0111a4ba
- function/view_image: View a local image file from the filesystem when visual inspection is needed. Use this for images...; args=[path]; hash=1908de9c0be7118a
- namespace/multi_agent_v1: Tools for spawning and managing sub-agents.; hash=f066a006a1fd060e
- namespace/multi_agent_v1: Tools for spawning and managing sub-agents.; hash=c041c3bd4f63167c
- function/close_agent
- function/resume_agent
- function/send_input
@@ -45,7 +45,7 @@ Settings:
## Window 2 (after request 1)
Settings: relative to window 1
tools (11; hash=64b8570765baaa12):
tools (11; hash=4a81d3515bbde49a):
~ namespace/mcp__discovery: Use these tools to exercise the rmcp test server.; hash=66db79539eaf6be3
- function/echo
- function/search_tools
@@ -16,7 +16,7 @@ Settings:
text: {"verbosity":"low"}
tool_choice: "auto"
-- request 1 (turn) --
00:additional_tools/developer (3; hash=2ee48a96ffd69424):
00:additional_tools/developer (3; hash=5133d148141b321a):
- namespace/functions; hash=81846de5bc4e7474
- custom/exec
- function/wait
@@ -24,7 +24,7 @@ Settings:
- function/request_user_input_async
- namespace/clock: Tools for reading and waiting on time.; hash=f402e8c5e9b5e317
- function/sleep
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=0c4edc9163319bad
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=bcd67096347e246f
- function/followup_task
- function/interrupt_agent
- function/list_agents
@@ -40,14 +40,14 @@ Settings:
User authorization and preferences persist across turns. Do not request permission again when the user h... any guidelines provided in skills or external files. [hash=911f91c92c72882c]
<OMITTED 155 LINES; ~4935 TOKENS; hash=5573869716192298>
<OMITTED 151 LINES; ~4851 TOKENS; hash=959762226fedfd0c>
## Decision Boundary
- Skill naming: If a plugin contributes skills, those skill entries are prefixed with plugin_name: in the Skills list.
- MCP naming: Plugin-provided MCP tools keep standard MCP identifiers such as mcp__server__tool; use tool provenance to tell which plugin they come from.
- Trigger rules: If the user explicitly names a plugin, prefer capabilities associated with that plugin for that turn.
- Relationship to capabilities: Plugins are not invoked directly. Use their underlying skills, MCP tools, and app tools to help solve the task.
- Relevance: Determine what a plugin can help with from explicit user mention or from the plugin-associa..., MCP tools, and apps exposed elsewhere in this turn. [hash=620459a4e7b590f0]
- Missing/blocked: If the user requests a plugin that does not have relevant callable capabilities for t..., say so briefly and continue with the best fallback. [hash=746b0469a8f6cb3e]
Discover and use relevant tools proactively when they may help carry out the requested task. For example...ng the user for information they may already contain. [hash=3af5c93cbc00b4c7]
When a plugin is available for a service and its tools are sufficient for completing the task, prefer th...er has an explicit preference, defer to that instead. [hash=936dbd7fe76c210f]
A user can also tag a plugin using `[@plugin-name](plugin://{{plugin_id}})` or an app using `[$app-name]... app, prefer its relevant capabilities for that turn. [hash=8730f69f41534f8f]
02:message/developer[2]:
[01] <permissions instructions>
@@ -33,14 +33,14 @@ Settings:
User authorization and preferences persist across turns. Do not request permission again when the user h... any guidelines provided in skills or external files. [hash=911f91c92c72882c]
<OMITTED 155 LINES; ~4935 TOKENS; hash=5573869716192298>
<OMITTED 151 LINES; ~4851 TOKENS; hash=959762226fedfd0c>
## Decision Boundary
- Skill naming: If a plugin contributes skills, those skill entries are prefixed with plugin_name: in the Skills list.
- MCP naming: Plugin-provided MCP tools keep standard MCP identifiers such as mcp__server__tool; use tool provenance to tell which plugin they come from.
- Trigger rules: If the user explicitly names a plugin, prefer capabilities associated with that plugin for that turn.
- Relationship to capabilities: Plugins are not invoked directly. Use their underlying skills, MCP tools, and app tools to help solve the task.
- Relevance: Determine what a plugin can help with from explicit user mention or from the plugin-associa..., MCP tools, and apps exposed elsewhere in this turn. [hash=620459a4e7b590f0]
- Missing/blocked: If the user requests a plugin that does not have relevant callable capabilities for t..., say so briefly and continue with the best fallback. [hash=746b0469a8f6cb3e]
Discover and use relevant tools proactively when they may help carry out the requested task. For example...ng the user for information they may already contain. [hash=3af5c93cbc00b4c7]
When a plugin is available for a service and its tools are sufficient for completing the task, prefer th...er has an explicit preference, defer to that instead. [hash=936dbd7fe76c210f]
A user can also tag a plugin using `[@plugin-name](plugin://{{plugin_id}})` or an app using `[$app-name]... app, prefer its relevant capabilities for that turn. [hash=8730f69f41534f8f]
02:message/developer[2]:
[01] <permissions instructions>
@@ -33,14 +33,14 @@ Settings:
User authorization and preferences persist across turns. Do not request permission again when the user h... any guidelines provided in skills or external files. [hash=911f91c92c72882c]
<OMITTED 155 LINES; ~4935 TOKENS; hash=5573869716192298>
<OMITTED 151 LINES; ~4851 TOKENS; hash=959762226fedfd0c>
## Decision Boundary
- Skill naming: If a plugin contributes skills, those skill entries are prefixed with plugin_name: in the Skills list.
- MCP naming: Plugin-provided MCP tools keep standard MCP identifiers such as mcp__server__tool; use tool provenance to tell which plugin they come from.
- Trigger rules: If the user explicitly names a plugin, prefer capabilities associated with that plugin for that turn.
- Relationship to capabilities: Plugins are not invoked directly. Use their underlying skills, MCP tools, and app tools to help solve the task.
- Relevance: Determine what a plugin can help with from explicit user mention or from the plugin-associa..., MCP tools, and apps exposed elsewhere in this turn. [hash=620459a4e7b590f0]
- Missing/blocked: If the user requests a plugin that does not have relevant callable capabilities for t..., say so briefly and continue with the best fallback. [hash=746b0469a8f6cb3e]
Discover and use relevant tools proactively when they may help carry out the requested task. For example...ng the user for information they may already contain. [hash=3af5c93cbc00b4c7]
When a plugin is available for a service and its tools are sufficient for completing the task, prefer th...er has an explicit preference, defer to that instead. [hash=936dbd7fe76c210f]
A user can also tag a plugin using `[@plugin-name](plugin://{{plugin_id}})` or an app using `[$app-name]... app, prefer its relevant capabilities for that turn. [hash=8730f69f41534f8f]
02:message/developer[2]:
[01] <permissions instructions>
@@ -19,7 +19,7 @@ Settings:
stream: true
text: {"verbosity":"low"}
tool_choice: "auto"
tools (14; hash=7bb8bff2b9bd81f8):
tools (14; hash=49d11ef121f73678):
- function/exec_command: Runs a command in a PTY, returning output or a session ID for ongoing interaction.; args=[cmd, environment_id, justification, login, max_output_tokens, prefix_rule, sandbox_permissions, shell, tty, workdir, yield_time_ms]; hash=b97d00b1a95ffa20
- function/write_stdin: Writes characters to an existing unified exec session and returns recent output.; args=[chars, max_output_tokens, session_id, yield_time_ms]; hash=7fb62d8a3e2df4e2
- function/list_mcp_resources: Lists resources provided by MCP servers. Resources allow servers to share data that provides cont...; args=[cursor, environment_id, server]; hash=54340f513f7e4f23
@@ -28,7 +28,7 @@ Settings:
- function/request_user_input: Request user input for one to three short questions and wait for the response. This tool is only ...; args=[questions]; hash=64a729b25c424fc6
- custom/apply_patch: The `apply_patch` tool can be used to edit files. This is a FREEFORM tool, so do not wrap the pat...; hash=6673c610c90583ae
- function/view_image: View a local image file from the filesystem when visual inspection is needed. Use this for images...; args=[environment_id, path]; hash=919f5105a7799bf4
- namespace/multi_agent_v1: Tools for spawning and managing sub-agents.; hash=f066a006a1fd060e
- namespace/multi_agent_v1: Tools for spawning and managing sub-agents.; hash=c041c3bd4f63167c
- function/close_agent
- function/resume_agent
- function/send_input
@@ -6,13 +6,13 @@ Scenario: A worker inherits the parent prefix, then receives its own window hint
## Window 1
-- request 1 (turn) --
00:additional_tools/developer (2; hash=9fb89f63da333416):
00:additional_tools/developer (2; hash=2c5684dede3a1a9c):
- namespace/functions; hash=c0e72eb3b93fb0da
- custom/exec
- function/wait
- function/request_user_input
- function/new_context
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=0271b89d55152d99
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=a097b9364047b8bb
- function/followup_task
- function/interrupt_agent
- function/list_agents
@@ -60,13 +60,13 @@ Scenario: A worker inherits the parent prefix, then receives its own window hint
## Window 2 (after request 3: input diverged at item 13)
-- request 4 (turn) --
00:additional_tools/developer (2; hash=9fb89f63da333416):
00:additional_tools/developer (2; hash=2c5684dede3a1a9c):
- namespace/functions; hash=c0e72eb3b93fb0da
- custom/exec
- function/wait
- function/request_user_input
- function/new_context
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=0271b89d55152d99
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=a097b9364047b8bb
- function/followup_task
- function/interrupt_agent
- function/list_agents
@@ -16,7 +16,7 @@ Settings:
text: {"verbosity":"low"}
tool_choice: "auto"
-- request 1 (turn) --
00:additional_tools/developer (3; hash=85dc7ee74cc74b50):
00:additional_tools/developer (3; hash=f98b02bc44103876):
- namespace/functions; hash=a1243ae7ec163900
- custom/exec
- function/wait
@@ -25,7 +25,7 @@ Settings:
- function/new_context
- namespace/clock: Tools for reading and waiting on time.; hash=f402e8c5e9b5e317
- function/sleep
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=0271b89d55152d99
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=a097b9364047b8bb
- function/followup_task
- function/interrupt_agent
- function/list_agents
@@ -41,14 +41,14 @@ Settings:
User authorization and preferences persist across turns. Do not request permission again when the user h... any guidelines provided in skills or external files. [hash=911f91c92c72882c]
<OMITTED 155 LINES; ~4935 TOKENS; hash=5573869716192298>
<OMITTED 151 LINES; ~4851 TOKENS; hash=959762226fedfd0c>
## Decision Boundary
- Skill naming: If a plugin contributes skills, those skill entries are prefixed with plugin_name: in the Skills list.
- MCP naming: Plugin-provided MCP tools keep standard MCP identifiers such as mcp__server__tool; use tool provenance to tell which plugin they come from.
- Trigger rules: If the user explicitly names a plugin, prefer capabilities associated with that plugin for that turn.
- Relationship to capabilities: Plugins are not invoked directly. Use their underlying skills, MCP tools, and app tools to help solve the task.
- Relevance: Determine what a plugin can help with from explicit user mention or from the plugin-associa..., MCP tools, and apps exposed elsewhere in this turn. [hash=620459a4e7b590f0]
- Missing/blocked: If the user requests a plugin that does not have relevant callable capabilities for t..., say so briefly and continue with the best fallback. [hash=746b0469a8f6cb3e]
Discover and use relevant tools proactively when they may help carry out the requested task. For example...ng the user for information they may already contain. [hash=3af5c93cbc00b4c7]
When a plugin is available for a service and its tools are sufficient for completing the task, prefer th...er has an explicit preference, defer to that instead. [hash=936dbd7fe76c210f]
A user can also tag a plugin using `[@plugin-name](plugin://{{plugin_id}})` or an app using `[$app-name]... app, prefer its relevant capabilities for that turn. [hash=8730f69f41534f8f]
02:message/developer[2]:
[01] <permissions instructions>
@@ -6,7 +6,7 @@ Scenario: With retention disabled, the client can still send retain=true. Tool o
## Window 1
-- request 1 (turn) --
00:additional_tools/developer (3; hash=f2e4edd02b6f9300):
00:additional_tools/developer (3; hash=a7c46a0959d54866):
- namespace/functions; hash=81846de5bc4e7474
- custom/exec
- function/wait
@@ -14,7 +14,7 @@ Scenario: With retention disabled, the client can still send retain=true. Tool o
- function/request_user_input_async
- namespace/clock: Tools for reading and waiting on time.; hash=f402e8c5e9b5e317
- function/sleep
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=0271b89d55152d99
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=a097b9364047b8bb
- function/followup_task
- function/interrupt_agent
- function/list_agents
@@ -45,7 +45,7 @@ Scenario: With retention disabled, the client can still send retain=true. Tool o
## Window 2 (after request 3: input diverged at item 11)
-- request 4 (compaction) --
00:additional_tools/developer (3; hash=f2e4edd02b6f9300):
00:additional_tools/developer (3; hash=a7c46a0959d54866):
- namespace/functions; hash=81846de5bc4e7474
- custom/exec
- function/wait
@@ -53,7 +53,7 @@ Scenario: With retention disabled, the client can still send retain=true. Tool o
- function/request_user_input_async
- namespace/clock: Tools for reading and waiting on time.; hash=f402e8c5e9b5e317
- function/sleep
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=0271b89d55152d99
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=a097b9364047b8bb
- function/followup_task
- function/interrupt_agent
- function/list_agents
@@ -84,7 +84,7 @@ Scenario: With retention disabled, the client can still send retain=true. Tool o
## Window 3 (after request 4: input diverged at item 06)
-- request 5 (turn) --
00:additional_tools/developer (3; hash=f2e4edd02b6f9300):
00:additional_tools/developer (3; hash=a7c46a0959d54866):
- namespace/functions; hash=81846de5bc4e7474
- custom/exec
- function/wait
@@ -92,7 +92,7 @@ Scenario: With retention disabled, the client can still send retain=true. Tool o
- function/request_user_input_async
- namespace/clock: Tools for reading and waiting on time.; hash=f402e8c5e9b5e317
- function/sleep
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=0271b89d55152d99
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=a097b9364047b8bb
- function/followup_task
- function/interrupt_agent
- function/list_agents
@@ -6,7 +6,7 @@ Scenario: Retained thread instructions stop emergency tool-output trimming. Comp
## Window 1
-- request 1 (turn) --
00:additional_tools/developer (3; hash=f2e4edd02b6f9300):
00:additional_tools/developer (3; hash=a7c46a0959d54866):
- namespace/functions; hash=81846de5bc4e7474
- custom/exec
- function/wait
@@ -14,7 +14,7 @@ Scenario: Retained thread instructions stop emergency tool-output trimming. Comp
- function/request_user_input_async
- namespace/clock: Tools for reading and waiting on time.; hash=f402e8c5e9b5e317
- function/sleep
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=0271b89d55152d99
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=a097b9364047b8bb
- function/followup_task
- function/interrupt_agent
- function/list_agents
@@ -48,7 +48,7 @@ Scenario: Retained thread instructions stop emergency tool-output trimming. Comp
## Window 2 (after request 4: input diverged at item 06)
-- request 5 (turn) --
00:additional_tools/developer (3; hash=f2e4edd02b6f9300):
00:additional_tools/developer (3; hash=a7c46a0959d54866):
- namespace/functions; hash=81846de5bc4e7474
- custom/exec
- function/wait
@@ -56,7 +56,7 @@ Scenario: Retained thread instructions stop emergency tool-output trimming. Comp
- function/request_user_input_async
- namespace/clock: Tools for reading and waiting on time.; hash=f402e8c5e9b5e317
- function/sleep
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=0271b89d55152d99
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=a097b9364047b8bb
- function/followup_task
- function/interrupt_agent
- function/list_agents
@@ -1,6 +1,6 @@
---
source: core/tests/suite/scenarios_skill_catalog_dedup.rs
expression: "context_snapshot::format_request_history_snapshot(\"Cloud skills replace duplicate executor listings before budgeting; unique executor skills retain their descriptions and e1 aliases across turns.\",\n&requests, &ContextSnapshotOptions::default().include_request_settings(),)"
expression: "context_snapshot::format_request_history_snapshot(if omit_descriptions\n{\n \"Cloud and executor skill catalogs retain names and aliases across turns with descriptions omitted.\"\n} else\n{\n \"Cloud skills replace duplicate executor listings before budgeting; unique executor skills retain their descriptions and e1 aliases across turns.\"\n}, &requests, &ContextSnapshotOptions::default().include_request_settings(),)"
---
Scenario: Cloud skills replace duplicate executor listings before budgeting; unique executor skills retain their descriptions and e1 aliases across turns.
@@ -16,7 +16,7 @@ Settings:
text: {"verbosity":"low"}
tool_choice: "auto"
-- request 1 (turn) --
00:additional_tools/developer (3; hash=aa1c8d2cbd56c247):
00:additional_tools/developer (3; hash=39cb3ee73f2e1987):
- namespace/functions; hash=c4874773c24012ab
- custom/exec
- function/wait
@@ -24,7 +24,7 @@ Settings:
- function/request_user_input_async
- namespace/clock: Tools for reading and waiting on time.; hash=f402e8c5e9b5e317
- function/sleep
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=0271b89d55152d99
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=a097b9364047b8bb
- function/followup_task
- function/interrupt_agent
- function/list_agents
@@ -40,14 +40,14 @@ Settings:
User authorization and preferences persist across turns. Do not request permission again when the user h... any guidelines provided in skills or external files. [hash=911f91c92c72882c]
<OMITTED 155 LINES; ~4935 TOKENS; hash=5573869716192298>
<OMITTED 151 LINES; ~4851 TOKENS; hash=959762226fedfd0c>
## Decision Boundary
- Skill naming: If a plugin contributes skills, those skill entries are prefixed with plugin_name: in the Skills list.
- MCP naming: Plugin-provided MCP tools keep standard MCP identifiers such as mcp__server__tool; use tool provenance to tell which plugin they come from.
- Trigger rules: If the user explicitly names a plugin, prefer capabilities associated with that plugin for that turn.
- Relationship to capabilities: Plugins are not invoked directly. Use their underlying skills, MCP tools, and app tools to help solve the task.
- Relevance: Determine what a plugin can help with from explicit user mention or from the plugin-associa..., MCP tools, and apps exposed elsewhere in this turn. [hash=620459a4e7b590f0]
- Missing/blocked: If the user requests a plugin that does not have relevant callable capabilities for t..., say so briefly and continue with the best fallback. [hash=746b0469a8f6cb3e]
Discover and use relevant tools proactively when they may help carry out the requested task. For example...ng the user for information they may already contain. [hash=3af5c93cbc00b4c7]
When a plugin is available for a service and its tools are sufficient for completing the task, prefer th...er has an explicit preference, defer to that instead. [hash=936dbd7fe76c210f]
A user can also tag a plugin using `[@plugin-name](plugin://{{plugin_id}})` or an app using `[$app-name]... app, prefer its relevant capabilities for that turn. [hash=8730f69f41534f8f]
02:message/developer[4]:
[01] <permissions instructions>
@@ -16,7 +16,7 @@ Settings:
text: {"verbosity":"low"}
tool_choice: "auto"
-- request 1 (turn) --
00:additional_tools/developer (3; hash=aa1c8d2cbd56c247):
00:additional_tools/developer (3; hash=39cb3ee73f2e1987):
- namespace/functions; hash=c4874773c24012ab
- custom/exec
- function/wait
@@ -24,7 +24,7 @@ Settings:
- function/request_user_input_async
- namespace/clock: Tools for reading and waiting on time.; hash=f402e8c5e9b5e317
- function/sleep
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=0271b89d55152d99
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=a097b9364047b8bb
- function/followup_task
- function/interrupt_agent
- function/list_agents
@@ -40,14 +40,14 @@ Settings:
User authorization and preferences persist across turns. Do not request permission again when the user h... any guidelines provided in skills or external files. [hash=911f91c92c72882c]
<OMITTED 155 LINES; ~4935 TOKENS; hash=5573869716192298>
<OMITTED 151 LINES; ~4851 TOKENS; hash=959762226fedfd0c>
## Decision Boundary
- Skill naming: If a plugin contributes skills, those skill entries are prefixed with plugin_name: in the Skills list.
- MCP naming: Plugin-provided MCP tools keep standard MCP identifiers such as mcp__server__tool; use tool provenance to tell which plugin they come from.
- Trigger rules: If the user explicitly names a plugin, prefer capabilities associated with that plugin for that turn.
- Relationship to capabilities: Plugins are not invoked directly. Use their underlying skills, MCP tools, and app tools to help solve the task.
- Relevance: Determine what a plugin can help with from explicit user mention or from the plugin-associa..., MCP tools, and apps exposed elsewhere in this turn. [hash=620459a4e7b590f0]
- Missing/blocked: If the user requests a plugin that does not have relevant callable capabilities for t..., say so briefly and continue with the best fallback. [hash=746b0469a8f6cb3e]
Discover and use relevant tools proactively when they may help carry out the requested task. For example...ng the user for information they may already contain. [hash=3af5c93cbc00b4c7]
When a plugin is available for a service and its tools are sufficient for completing the task, prefer th...er has an explicit preference, defer to that instead. [hash=936dbd7fe76c210f]
A user can also tag a plugin using `[@plugin-name](plugin://{{plugin_id}})` or an app using `[$app-name]... app, prefer its relevant capabilities for that turn. [hash=8730f69f41534f8f]
02:message/developer[4]:
[01] <permissions instructions>
@@ -16,7 +16,7 @@ Settings:
text: {"verbosity":"low"}
tool_choice: "auto"
-- request 1 (turn) --
00:additional_tools/developer (3; hash=f2e4edd02b6f9300):
00:additional_tools/developer (3; hash=a7c46a0959d54866):
- namespace/functions; hash=81846de5bc4e7474
- custom/exec
- function/wait
@@ -24,7 +24,7 @@ Settings:
- function/request_user_input_async
- namespace/clock: Tools for reading and waiting on time.; hash=f402e8c5e9b5e317
- function/sleep
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=0271b89d55152d99
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=a097b9364047b8bb
- function/followup_task
- function/interrupt_agent
- function/list_agents
@@ -40,14 +40,14 @@ Settings:
User authorization and preferences persist across turns. Do not request permission again when the user h... any guidelines provided in skills or external files. [hash=911f91c92c72882c]
<OMITTED 155 LINES; ~4935 TOKENS; hash=5573869716192298>
<OMITTED 151 LINES; ~4851 TOKENS; hash=959762226fedfd0c>
## Decision Boundary
- Skill naming: If a plugin contributes skills, those skill entries are prefixed with plugin_name: in the Skills list.
- MCP naming: Plugin-provided MCP tools keep standard MCP identifiers such as mcp__server__tool; use tool provenance to tell which plugin they come from.
- Trigger rules: If the user explicitly names a plugin, prefer capabilities associated with that plugin for that turn.
- Relationship to capabilities: Plugins are not invoked directly. Use their underlying skills, MCP tools, and app tools to help solve the task.
- Relevance: Determine what a plugin can help with from explicit user mention or from the plugin-associa..., MCP tools, and apps exposed elsewhere in this turn. [hash=620459a4e7b590f0]
- Missing/blocked: If the user requests a plugin that does not have relevant callable capabilities for t..., say so briefly and continue with the best fallback. [hash=746b0469a8f6cb3e]
Discover and use relevant tools proactively when they may help carry out the requested task. For example...ng the user for information they may already contain. [hash=3af5c93cbc00b4c7]
When a plugin is available for a service and its tools are sufficient for completing the task, prefer th...er has an explicit preference, defer to that instead. [hash=936dbd7fe76c210f]
A user can also tag a plugin using `[@plugin-name](plugin://{{plugin_id}})` or an app using `[$app-name]... app, prefer its relevant capabilities for that turn. [hash=8730f69f41534f8f]
02:message/developer[3]:
[01] <skills_instructions>
@@ -160,7 +160,7 @@ Settings:
## Window 2 (after request 7: input diverged at item 02)
Settings: same as window 1
-- request 8 (turn) --
00:additional_tools/developer (3; hash=f2e4edd02b6f9300):
00:additional_tools/developer (3; hash=a7c46a0959d54866):
- namespace/functions; hash=81846de5bc4e7474
- custom/exec
- function/wait
@@ -168,7 +168,7 @@ Settings: same as window 1
- function/request_user_input_async
- namespace/clock: Tools for reading and waiting on time.; hash=f402e8c5e9b5e317
- function/sleep
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=0271b89d55152d99
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=a097b9364047b8bb
- function/followup_task
- function/interrupt_agent
- function/list_agents
@@ -184,14 +184,14 @@ Settings: same as window 1
User authorization and preferences persist across turns. Do not request permission again when the user h... any guidelines provided in skills or external files. [hash=911f91c92c72882c]
<OMITTED 155 LINES; ~4935 TOKENS; hash=5573869716192298>
<OMITTED 151 LINES; ~4851 TOKENS; hash=959762226fedfd0c>
## Decision Boundary
- Skill naming: If a plugin contributes skills, those skill entries are prefixed with plugin_name: in the Skills list.
- MCP naming: Plugin-provided MCP tools keep standard MCP identifiers such as mcp__server__tool; use tool provenance to tell which plugin they come from.
- Trigger rules: If the user explicitly names a plugin, prefer capabilities associated with that plugin for that turn.
- Relationship to capabilities: Plugins are not invoked directly. Use their underlying skills, MCP tools, and app tools to help solve the task.
- Relevance: Determine what a plugin can help with from explicit user mention or from the plugin-associa..., MCP tools, and apps exposed elsewhere in this turn. [hash=620459a4e7b590f0]
- Missing/blocked: If the user requests a plugin that does not have relevant callable capabilities for t..., say so briefly and continue with the best fallback. [hash=746b0469a8f6cb3e]
Discover and use relevant tools proactively when they may help carry out the requested task. For example...ng the user for information they may already contain. [hash=3af5c93cbc00b4c7]
When a plugin is available for a service and its tools are sufficient for completing the task, prefer th...er has an explicit preference, defer to that instead. [hash=936dbd7fe76c210f]
A user can also tag a plugin using `[@plugin-name](plugin://{{plugin_id}})` or an app using `[$app-name]... app, prefer its relevant capabilities for that turn. [hash=8730f69f41534f8f]
02:message/user:
Inspect the available skills.
@@ -295,7 +295,7 @@ Settings: same as window 1
## Window 3 (after request 11: input diverged at item 07)
Settings: same as window 1
-- request 12 (turn) --
00:additional_tools/developer (3; hash=f2e4edd02b6f9300):
00:additional_tools/developer (3; hash=a7c46a0959d54866):
- namespace/functions; hash=81846de5bc4e7474
- custom/exec
- function/wait
@@ -303,7 +303,7 @@ Settings: same as window 1
- function/request_user_input_async
- namespace/clock: Tools for reading and waiting on time.; hash=f402e8c5e9b5e317
- function/sleep
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=0271b89d55152d99
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=a097b9364047b8bb
- function/followup_task
- function/interrupt_agent
- function/list_agents
@@ -319,14 +319,14 @@ Settings: same as window 1
User authorization and preferences persist across turns. Do not request permission again when the user h... any guidelines provided in skills or external files. [hash=911f91c92c72882c]
<OMITTED 155 LINES; ~4935 TOKENS; hash=5573869716192298>
<OMITTED 151 LINES; ~4851 TOKENS; hash=959762226fedfd0c>
## Decision Boundary
- Skill naming: If a plugin contributes skills, those skill entries are prefixed with plugin_name: in the Skills list.
- MCP naming: Plugin-provided MCP tools keep standard MCP identifiers such as mcp__server__tool; use tool provenance to tell which plugin they come from.
- Trigger rules: If the user explicitly names a plugin, prefer capabilities associated with that plugin for that turn.
- Relationship to capabilities: Plugins are not invoked directly. Use their underlying skills, MCP tools, and app tools to help solve the task.
- Relevance: Determine what a plugin can help with from explicit user mention or from the plugin-associa..., MCP tools, and apps exposed elsewhere in this turn. [hash=620459a4e7b590f0]
- Missing/blocked: If the user requests a plugin that does not have relevant callable capabilities for t..., say so briefly and continue with the best fallback. [hash=746b0469a8f6cb3e]
Discover and use relevant tools proactively when they may help carry out the requested task. For example...ng the user for information they may already contain. [hash=3af5c93cbc00b4c7]
When a plugin is available for a service and its tools are sufficient for completing the task, prefer th...er has an explicit preference, defer to that instead. [hash=936dbd7fe76c210f]
A user can also tag a plugin using `[@plugin-name](plugin://{{plugin_id}})` or an app using `[$app-name]... app, prefer its relevant capabilities for that turn. [hash=8730f69f41534f8f]
02:message/user:
Inspect the available skills.
@@ -436,7 +436,7 @@ Settings: same as window 1
## Window 4 (after request 18: input diverged at item 10)
Settings: same as window 1
-- request 19 (turn) --
00:additional_tools/developer (3; hash=f2e4edd02b6f9300):
00:additional_tools/developer (3; hash=a7c46a0959d54866):
- namespace/functions; hash=81846de5bc4e7474
- custom/exec
- function/wait
@@ -444,7 +444,7 @@ Settings: same as window 1
- function/request_user_input_async
- namespace/clock: Tools for reading and waiting on time.; hash=f402e8c5e9b5e317
- function/sleep
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=0271b89d55152d99
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=a097b9364047b8bb
- function/followup_task
- function/interrupt_agent
- function/list_agents
@@ -460,14 +460,14 @@ Settings: same as window 1
User authorization and preferences persist across turns. Do not request permission again when the user h... any guidelines provided in skills or external files. [hash=911f91c92c72882c]
<OMITTED 155 LINES; ~4935 TOKENS; hash=5573869716192298>
<OMITTED 151 LINES; ~4851 TOKENS; hash=959762226fedfd0c>
## Decision Boundary
- Skill naming: If a plugin contributes skills, those skill entries are prefixed with plugin_name: in the Skills list.
- MCP naming: Plugin-provided MCP tools keep standard MCP identifiers such as mcp__server__tool; use tool provenance to tell which plugin they come from.
- Trigger rules: If the user explicitly names a plugin, prefer capabilities associated with that plugin for that turn.
- Relationship to capabilities: Plugins are not invoked directly. Use their underlying skills, MCP tools, and app tools to help solve the task.
- Relevance: Determine what a plugin can help with from explicit user mention or from the plugin-associa..., MCP tools, and apps exposed elsewhere in this turn. [hash=620459a4e7b590f0]
- Missing/blocked: If the user requests a plugin that does not have relevant callable capabilities for t..., say so briefly and continue with the best fallback. [hash=746b0469a8f6cb3e]
Discover and use relevant tools proactively when they may help carry out the requested task. For example...ng the user for information they may already contain. [hash=3af5c93cbc00b4c7]
When a plugin is available for a service and its tools are sufficient for completing the task, prefer th...er has an explicit preference, defer to that instead. [hash=936dbd7fe76c210f]
A user can also tag a plugin using `[@plugin-name](plugin://{{plugin_id}})` or an app using `[$app-name]... app, prefer its relevant capabilities for that turn. [hash=8730f69f41534f8f]
02:message/user:
Inspect the available skills.
@@ -6,7 +6,7 @@ Scenario: Astra reads Cloud instructions, discovers an omitted deployment instal
## Window 1
-- request 1 (request) --
00:additional_tools/developer (3; hash=e37d13847f694d9d):
00:additional_tools/developer (3; hash=8f90f1c7f4fce935):
- namespace/functions; hash=3eb85b1b397a4861
- custom/exec
- function/wait
@@ -14,7 +14,7 @@ Scenario: Astra reads Cloud instructions, discovers an omitted deployment instal
- function/request_user_input_async
- namespace/clock: Tools for reading and waiting on time.; hash=f402e8c5e9b5e317
- function/sleep
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=0271b89d55152d99
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=a097b9364047b8bb
- function/followup_task
- function/interrupt_agent
- function/list_agents
@@ -6,7 +6,7 @@ Scenario: A user requests the bundled statusline skill by name.
## Window 1
-- request 1 (turn) --
00:additional_tools/developer (3; hash=f2e4edd02b6f9300):
00:additional_tools/developer (3; hash=a7c46a0959d54866):
- namespace/functions; hash=81846de5bc4e7474
- custom/exec
- function/wait
@@ -14,7 +14,7 @@ Scenario: A user requests the bundled statusline skill by name.
- function/request_user_input_async
- namespace/clock: Tools for reading and waiting on time.; hash=f402e8c5e9b5e317
- function/sleep
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=0271b89d55152d99
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=a097b9364047b8bb
- function/followup_task
- function/interrupt_agent
- function/list_agents
@@ -24,7 +24,7 @@ Settings:
- function/request_user_input_async
- namespace/clock: Tools for reading and waiting on time.; hash=f402e8c5e9b5e317
- function/sleep
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=0271b89d55152d99
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=a097b9364047b8bb
- function/followup_task
- function/interrupt_agent
- function/list_agents
@@ -1,13 +1,12 @@
---
source: core/tests/suite/scenarios.rs
assertion_line: 723
expression: "context_snapshot::format_request_history_snapshot(\"Available file references pass through. A later request replaces missing files, preserves references on backend errors, rechecks each distinct ID once, and keeps canonical history intact.\",\n&requests, &ContextSnapshotOptions::default().rewrite_known_segments(),)"
---
Scenario: Available file references pass through. A later request replaces missing files, preserves references on backend errors, rechecks each distinct ID once, and keeps canonical history intact.
## Window 1
-- request 1 (turn) --
00:additional_tools/developer (3; hash=f2e4edd02b6f9300):
00:additional_tools/developer (3; hash=a7c46a0959d54866):
- namespace/functions; hash=81846de5bc4e7474
- custom/exec
- function/wait
@@ -15,7 +14,7 @@ Scenario: Available file references pass through. A later request replaces missi
- function/request_user_input_async
- namespace/clock: Tools for reading and waiting on time.; hash=f402e8c5e9b5e317
- function/sleep
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=0271b89d55152d99
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=a097b9364047b8bb
- function/followup_task
- function/interrupt_agent
- function/list_agents
@@ -43,7 +42,7 @@ Scenario: Available file references pass through. A later request replaces missi
## Window 2 (after request 1: input diverged at item 06)
-- request 2 (turn) --
00:additional_tools/developer (3; hash=f2e4edd02b6f9300):
00:additional_tools/developer (3; hash=a7c46a0959d54866):
- namespace/functions; hash=81846de5bc4e7474
- custom/exec
- function/wait
@@ -51,7 +50,7 @@ Scenario: Available file references pass through. A later request replaces missi
- function/request_user_input_async
- namespace/clock: Tools for reading and waiting on time.; hash=f402e8c5e9b5e317
- function/sleep
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=0271b89d55152d99
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=a097b9364047b8bb
- function/followup_task
- function/interrupt_agent
- function/list_agents
@@ -16,7 +16,7 @@ Settings:
text: {"verbosity":"low"}
tool_choice: "auto"
-- request 1 (turn; Initial policy A) --
00:additional_tools/developer (3; hash=f2e4edd02b6f9300):
00:additional_tools/developer (3; hash=a7c46a0959d54866):
- namespace/functions; hash=81846de5bc4e7474
- custom/exec
- function/wait
@@ -24,7 +24,7 @@ Settings:
- function/request_user_input_async
- namespace/clock: Tools for reading and waiting on time.; hash=f402e8c5e9b5e317
- function/sleep
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=0271b89d55152d99
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=a097b9364047b8bb
- function/followup_task
- function/interrupt_agent
- function/list_agents
@@ -40,14 +40,14 @@ Settings:
User authorization and preferences persist across turns. Do not request permission again when the user h... any guidelines provided in skills or external files. [hash=911f91c92c72882c]
<OMITTED 155 LINES; ~4935 TOKENS; hash=5573869716192298>
<OMITTED 151 LINES; ~4851 TOKENS; hash=959762226fedfd0c>
## Decision Boundary
- Skill naming: If a plugin contributes skills, those skill entries are prefixed with plugin_name: in the Skills list.
- MCP naming: Plugin-provided MCP tools keep standard MCP identifiers such as mcp__server__tool; use tool provenance to tell which plugin they come from.
- Trigger rules: If the user explicitly names a plugin, prefer capabilities associated with that plugin for that turn.
- Relationship to capabilities: Plugins are not invoked directly. Use their underlying skills, MCP tools, and app tools to help solve the task.
- Relevance: Determine what a plugin can help with from explicit user mention or from the plugin-associa..., MCP tools, and apps exposed elsewhere in this turn. [hash=620459a4e7b590f0]
- Missing/blocked: If the user requests a plugin that does not have relevant callable capabilities for t..., say so briefly and continue with the best fallback. [hash=746b0469a8f6cb3e]
Discover and use relevant tools proactively when they may help carry out the requested task. For example...ng the user for information they may already contain. [hash=3af5c93cbc00b4c7]
When a plugin is available for a service and its tools are sufficient for completing the task, prefer th...er has an explicit preference, defer to that instead. [hash=936dbd7fe76c210f]
A user can also tag a plugin using `[@plugin-name](plugin://{{plugin_id}})` or an app using `[$app-name]... app, prefer its relevant capabilities for that turn. [hash=8730f69f41534f8f]
02:message/developer:
<scenario_instructions>scenario_prefix: A</scenario_instructions>
@@ -132,7 +132,7 @@ Settings:
## Window 2 (after request 4: input diverged at item 02)
Settings: unavailable (items only)
-- request 5 (items; Stored replacement before next turn) --
00:additional_tools/developer (3; hash=f2e4edd02b6f9300):
00:additional_tools/developer (3; hash=a7c46a0959d54866):
- namespace/functions; hash=81846de5bc4e7474
- custom/exec
- function/wait
@@ -140,7 +140,7 @@ Settings: unavailable (items only)
- function/request_user_input_async
- namespace/clock: Tools for reading and waiting on time.; hash=f402e8c5e9b5e317
- function/sleep
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=0271b89d55152d99
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=a097b9364047b8bb
- function/followup_task
- function/interrupt_agent
- function/list_agents
@@ -156,14 +156,14 @@ Settings: unavailable (items only)
User authorization and preferences persist across turns. Do not request permission again when the user h... any guidelines provided in skills or external files. [hash=911f91c92c72882c]
<OMITTED 155 LINES; ~4935 TOKENS; hash=5573869716192298>
<OMITTED 151 LINES; ~4851 TOKENS; hash=959762226fedfd0c>
## Decision Boundary
- Skill naming: If a plugin contributes skills, those skill entries are prefixed with plugin_name: in the Skills list.
- MCP naming: Plugin-provided MCP tools keep standard MCP identifiers such as mcp__server__tool; use tool provenance to tell which plugin they come from.
- Trigger rules: If the user explicitly names a plugin, prefer capabilities associated with that plugin for that turn.
- Relationship to capabilities: Plugins are not invoked directly. Use their underlying skills, MCP tools, and app tools to help solve the task.
- Relevance: Determine what a plugin can help with from explicit user mention or from the plugin-associa..., MCP tools, and apps exposed elsewhere in this turn. [hash=620459a4e7b590f0]
- Missing/blocked: If the user requests a plugin that does not have relevant callable capabilities for t..., say so briefly and continue with the best fallback. [hash=746b0469a8f6cb3e]
Discover and use relevant tools proactively when they may help carry out the requested task. For example...ng the user for information they may already contain. [hash=3af5c93cbc00b4c7]
When a plugin is available for a service and its tools are sufficient for completing the task, prefer th...er has an explicit preference, defer to that instead. [hash=936dbd7fe76c210f]
A user can also tag a plugin using `[@plugin-name](plugin://{{plugin_id}})` or an app using `[$app-name]... app, prefer its relevant capabilities for that turn. [hash=8730f69f41534f8f]
02:message/developer:
<scenario_instructions>scenario_prefix: C</scenario_instructions>
@@ -229,7 +229,7 @@ Settings: unavailable (items only)
## Window 3 (after request 5)
Settings: same as window 1
-- request 6 (turn; Unchanged policy C after compaction) --
00:additional_tools/developer (3; hash=f2e4edd02b6f9300):
00:additional_tools/developer (3; hash=a7c46a0959d54866):
- namespace/functions; hash=81846de5bc4e7474
- custom/exec
- function/wait
@@ -237,7 +237,7 @@ Settings: same as window 1
- function/request_user_input_async
- namespace/clock: Tools for reading and waiting on time.; hash=f402e8c5e9b5e317
- function/sleep
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=0271b89d55152d99
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=a097b9364047b8bb
- function/followup_task
- function/interrupt_agent
- function/list_agents
@@ -253,14 +253,14 @@ Settings: same as window 1
User authorization and preferences persist across turns. Do not request permission again when the user h... any guidelines provided in skills or external files. [hash=911f91c92c72882c]
<OMITTED 155 LINES; ~4935 TOKENS; hash=5573869716192298>
<OMITTED 151 LINES; ~4851 TOKENS; hash=959762226fedfd0c>
## Decision Boundary
- Skill naming: If a plugin contributes skills, those skill entries are prefixed with plugin_name: in the Skills list.
- MCP naming: Plugin-provided MCP tools keep standard MCP identifiers such as mcp__server__tool; use tool provenance to tell which plugin they come from.
- Trigger rules: If the user explicitly names a plugin, prefer capabilities associated with that plugin for that turn.
- Relationship to capabilities: Plugins are not invoked directly. Use their underlying skills, MCP tools, and app tools to help solve the task.
- Relevance: Determine what a plugin can help with from explicit user mention or from the plugin-associa..., MCP tools, and apps exposed elsewhere in this turn. [hash=620459a4e7b590f0]
- Missing/blocked: If the user requests a plugin that does not have relevant callable capabilities for t..., say so briefly and continue with the best fallback. [hash=746b0469a8f6cb3e]
Discover and use relevant tools proactively when they may help carry out the requested task. For example...ng the user for information they may already contain. [hash=3af5c93cbc00b4c7]
When a plugin is available for a service and its tools are sufficient for completing the task, prefer th...er has an explicit preference, defer to that instead. [hash=936dbd7fe76c210f]
A user can also tag a plugin using `[@plugin-name](plugin://{{plugin_id}})` or an app using `[$app-name]... app, prefer its relevant capabilities for that turn. [hash=8730f69f41534f8f]
02:message/developer:
<scenario_instructions>scenario_prefix: C</scenario_instructions>
@@ -16,7 +16,7 @@ Settings:
text: {"verbosity":"low"}
tool_choice: "auto"
-- request 1 (turn; Initial policy A) --
00:additional_tools/developer (3; hash=f2e4edd02b6f9300):
00:additional_tools/developer (3; hash=a7c46a0959d54866):
- namespace/functions; hash=81846de5bc4e7474
- custom/exec
- function/wait
@@ -24,7 +24,7 @@ Settings:
- function/request_user_input_async
- namespace/clock: Tools for reading and waiting on time.; hash=f402e8c5e9b5e317
- function/sleep
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=0271b89d55152d99
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=a097b9364047b8bb
- function/followup_task
- function/interrupt_agent
- function/list_agents
@@ -40,14 +40,14 @@ Settings:
User authorization and preferences persist across turns. Do not request permission again when the user h... any guidelines provided in skills or external files. [hash=911f91c92c72882c]
<OMITTED 155 LINES; ~4935 TOKENS; hash=5573869716192298>
<OMITTED 151 LINES; ~4851 TOKENS; hash=959762226fedfd0c>
## Decision Boundary
- Skill naming: If a plugin contributes skills, those skill entries are prefixed with plugin_name: in the Skills list.
- MCP naming: Plugin-provided MCP tools keep standard MCP identifiers such as mcp__server__tool; use tool provenance to tell which plugin they come from.
- Trigger rules: If the user explicitly names a plugin, prefer capabilities associated with that plugin for that turn.
- Relationship to capabilities: Plugins are not invoked directly. Use their underlying skills, MCP tools, and app tools to help solve the task.
- Relevance: Determine what a plugin can help with from explicit user mention or from the plugin-associa..., MCP tools, and apps exposed elsewhere in this turn. [hash=620459a4e7b590f0]
- Missing/blocked: If the user requests a plugin that does not have relevant callable capabilities for t..., say so briefly and continue with the best fallback. [hash=746b0469a8f6cb3e]
Discover and use relevant tools proactively when they may help carry out the requested task. For example...ng the user for information they may already contain. [hash=3af5c93cbc00b4c7]
When a plugin is available for a service and its tools are sufficient for completing the task, prefer th...er has an explicit preference, defer to that instead. [hash=936dbd7fe76c210f]
A user can also tag a plugin using `[@plugin-name](plugin://{{plugin_id}})` or an app using `[$app-name]... app, prefer its relevant capabilities for that turn. [hash=8730f69f41534f8f]
02:message/developer:
<scenario_instructions>scenario_prefix: A</scenario_instructions>
@@ -123,7 +123,7 @@ Settings:
## Window 2 (after request 4: input diverged at item 02)
Settings: unavailable (items only)
-- request 5 (items; Stored replacement before next turn) --
00:additional_tools/developer (3; hash=f2e4edd02b6f9300):
00:additional_tools/developer (3; hash=a7c46a0959d54866):
- namespace/functions; hash=81846de5bc4e7474
- custom/exec
- function/wait
@@ -131,7 +131,7 @@ Settings: unavailable (items only)
- function/request_user_input_async
- namespace/clock: Tools for reading and waiting on time.; hash=f402e8c5e9b5e317
- function/sleep
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=0271b89d55152d99
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=a097b9364047b8bb
- function/followup_task
- function/interrupt_agent
- function/list_agents
@@ -147,14 +147,14 @@ Settings: unavailable (items only)
User authorization and preferences persist across turns. Do not request permission again when the user h... any guidelines provided in skills or external files. [hash=911f91c92c72882c]
<OMITTED 155 LINES; ~4935 TOKENS; hash=5573869716192298>
<OMITTED 151 LINES; ~4851 TOKENS; hash=959762226fedfd0c>
## Decision Boundary
- Skill naming: If a plugin contributes skills, those skill entries are prefixed with plugin_name: in the Skills list.
- MCP naming: Plugin-provided MCP tools keep standard MCP identifiers such as mcp__server__tool; use tool provenance to tell which plugin they come from.
- Trigger rules: If the user explicitly names a plugin, prefer capabilities associated with that plugin for that turn.
- Relationship to capabilities: Plugins are not invoked directly. Use their underlying skills, MCP tools, and app tools to help solve the task.
- Relevance: Determine what a plugin can help with from explicit user mention or from the plugin-associa..., MCP tools, and apps exposed elsewhere in this turn. [hash=620459a4e7b590f0]
- Missing/blocked: If the user requests a plugin that does not have relevant callable capabilities for t..., say so briefly and continue with the best fallback. [hash=746b0469a8f6cb3e]
Discover and use relevant tools proactively when they may help carry out the requested task. For example...ng the user for information they may already contain. [hash=3af5c93cbc00b4c7]
When a plugin is available for a service and its tools are sufficient for completing the task, prefer th...er has an explicit preference, defer to that instead. [hash=936dbd7fe76c210f]
A user can also tag a plugin using `[@plugin-name](plugin://{{plugin_id}})` or an app using `[$app-name]... app, prefer its relevant capabilities for that turn. [hash=8730f69f41534f8f]
02:message/developer:
<scenario_instructions>scenario_prefix: C</scenario_instructions>
@@ -218,7 +218,7 @@ Settings: unavailable (items only)
## Window 3 (after request 5)
Settings: same as window 1
-- request 6 (turn; Unchanged policy C after compaction) --
00:additional_tools/developer (3; hash=f2e4edd02b6f9300):
00:additional_tools/developer (3; hash=a7c46a0959d54866):
- namespace/functions; hash=81846de5bc4e7474
- custom/exec
- function/wait
@@ -226,7 +226,7 @@ Settings: same as window 1
- function/request_user_input_async
- namespace/clock: Tools for reading and waiting on time.; hash=f402e8c5e9b5e317
- function/sleep
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=0271b89d55152d99
- namespace/collaboration: Tools for spawning and managing sub-agents.; hash=a097b9364047b8bb
- function/followup_task
- function/interrupt_agent
- function/list_agents
@@ -242,14 +242,14 @@ Settings: same as window 1
User authorization and preferences persist across turns. Do not request permission again when the user h... any guidelines provided in skills or external files. [hash=911f91c92c72882c]
<OMITTED 155 LINES; ~4935 TOKENS; hash=5573869716192298>
<OMITTED 151 LINES; ~4851 TOKENS; hash=959762226fedfd0c>
## Decision Boundary
- Skill naming: If a plugin contributes skills, those skill entries are prefixed with plugin_name: in the Skills list.
- MCP naming: Plugin-provided MCP tools keep standard MCP identifiers such as mcp__server__tool; use tool provenance to tell which plugin they come from.
- Trigger rules: If the user explicitly names a plugin, prefer capabilities associated with that plugin for that turn.
- Relationship to capabilities: Plugins are not invoked directly. Use their underlying skills, MCP tools, and app tools to help solve the task.
- Relevance: Determine what a plugin can help with from explicit user mention or from the plugin-associa..., MCP tools, and apps exposed elsewhere in this turn. [hash=620459a4e7b590f0]
- Missing/blocked: If the user requests a plugin that does not have relevant callable capabilities for t..., say so briefly and continue with the best fallback. [hash=746b0469a8f6cb3e]
Discover and use relevant tools proactively when they may help carry out the requested task. For example...ng the user for information they may already contain. [hash=3af5c93cbc00b4c7]
When a plugin is available for a service and its tools are sufficient for completing the task, prefer th...er has an explicit preference, defer to that instead. [hash=936dbd7fe76c210f]
A user can also tag a plugin using `[@plugin-name](plugin://{{plugin_id}})` or an app using `[$app-name]... app, prefer its relevant capabilities for that turn. [hash=8730f69f41534f8f]
02:message/developer:
<scenario_instructions>scenario_prefix: C</scenario_instructions>
+2
View File
@@ -95,6 +95,8 @@ pub const AMAZON_BEDROCK_GPT_5_5_MODEL_ID: &str = "openai.gpt-5.5";
pub const AMAZON_BEDROCK_GPT_5_6_SOL_MODEL_ID: &str = "openai.gpt-5.6-sol";
pub const AMAZON_BEDROCK_GPT_6_1_SOL_MODEL_ID: &str = "openai.gpt-6.1-sol";
pub const AMAZON_BEDROCK_GPT_6_1_TERRA_MODEL_ID: &str = "openai.gpt-6.1-terra";
pub const AMAZON_BEDROCK_GPT_6_2_ASTRA_MODEL_ID: &str = "openai.gpt-6.2-astra";
pub const AMAZON_BEDROCK_GPT_6_2_SOL_MODEL_ID: &str = "openai.gpt-6.2-sol";
pub const AMAZON_BEDROCK_GPT_6_SOL_MODEL_ID: &str = "openai.gpt-6-sol";
pub const AMAZON_BEDROCK_GPT_6_LUNA_MODEL_ID: &str = "openai.gpt-6-luna";
pub const AMAZON_BEDROCK_GPT_6_ASTRA_MODEL_ID: &str = "openai.gpt-6-astra";
@@ -4,6 +4,8 @@ use codex_model_provider_info::AMAZON_BEDROCK_GPT_5_6_SOL_MODEL_ID;
use codex_model_provider_info::AMAZON_BEDROCK_GPT_5_6_TERRA_MODEL_ID;
use codex_model_provider_info::AMAZON_BEDROCK_GPT_6_1_SOL_MODEL_ID;
use codex_model_provider_info::AMAZON_BEDROCK_GPT_6_1_TERRA_MODEL_ID;
use codex_model_provider_info::AMAZON_BEDROCK_GPT_6_2_ASTRA_MODEL_ID;
use codex_model_provider_info::AMAZON_BEDROCK_GPT_6_2_SOL_MODEL_ID;
use codex_model_provider_info::AMAZON_BEDROCK_GPT_6_ASTRA_MODEL_ID;
use codex_model_provider_info::AMAZON_BEDROCK_GPT_6_LUNA_MODEL_ID;
use codex_model_provider_info::AMAZON_BEDROCK_GPT_6_SOL_MODEL_ID;
@@ -22,6 +24,8 @@ const GPT_5_6_TERRA_OPENAI_MODEL_ID: &str = "gpt-5.6-terra";
const GPT_5_6_LUNA_OPENAI_MODEL_ID: &str = "gpt-5.6-luna";
const GPT_6_1_SOL_OPENAI_MODEL_ID: &str = "gpt-6.1-sol";
const GPT_6_1_TERRA_OPENAI_MODEL_ID: &str = "gpt-6.1-terra";
const GPT_6_2_ASTRA_OPENAI_MODEL_ID: &str = "gpt-6.2-astra";
const GPT_6_2_SOL_OPENAI_MODEL_ID: &str = "gpt-6.2-sol";
const GPT_6_SOL_OPENAI_MODEL_ID: &str = "gpt-6-sol";
const GPT_6_LUNA_OPENAI_MODEL_ID: &str = "gpt-6-luna";
const GPT_6_ASTRA_OPENAI_MODEL_ID: &str = "gpt-6-astra";
@@ -31,15 +35,15 @@ pub(crate) fn static_model_catalog() -> ModelsResponse {
let catalog = ModelsResponse {
models: vec![
bedrock_model(
bundled_openai_model(GPT_6_1_SOL_OPENAI_MODEL_ID),
AMAZON_BEDROCK_GPT_6_1_SOL_MODEL_ID,
"GPT-6.1 Sol",
bundled_openai_model(GPT_6_2_SOL_OPENAI_MODEL_ID),
AMAZON_BEDROCK_GPT_6_2_SOL_MODEL_ID,
"GPT-6.2 Sol",
/*priority*/ 0,
),
bedrock_model(
bundled_openai_model(GPT_6_ASTRA_OPENAI_MODEL_ID),
AMAZON_BEDROCK_GPT_6_ASTRA_MODEL_ID,
"GPT-6-Astra",
bundled_openai_model(GPT_6_2_ASTRA_OPENAI_MODEL_ID),
AMAZON_BEDROCK_GPT_6_2_ASTRA_MODEL_ID,
"GPT-6.2 Astra",
/*priority*/ 1,
),
bedrock_model(
@@ -48,41 +52,53 @@ pub(crate) fn static_model_catalog() -> ModelsResponse {
"GPT-6.1 Terra",
/*priority*/ 2,
),
bedrock_model(
bundled_openai_model(GPT_6_1_SOL_OPENAI_MODEL_ID),
AMAZON_BEDROCK_GPT_6_1_SOL_MODEL_ID,
"GPT-6.1 Sol",
/*priority*/ 3,
),
bedrock_model(
bundled_openai_model(GPT_6_ASTRA_OPENAI_MODEL_ID),
AMAZON_BEDROCK_GPT_6_ASTRA_MODEL_ID,
"GPT-6-Astra",
/*priority*/ 4,
),
bedrock_model(
bundled_openai_model(GPT_6_SOL_OPENAI_MODEL_ID),
AMAZON_BEDROCK_GPT_6_SOL_MODEL_ID,
"GPT-6 Sol",
/*priority*/ 3,
/*priority*/ 5,
),
bedrock_model(
bundled_openai_model(GPT_6_LUNA_OPENAI_MODEL_ID),
AMAZON_BEDROCK_GPT_6_LUNA_MODEL_ID,
"GPT-6 Luna",
/*priority*/ 4,
/*priority*/ 6,
),
bedrock_model(
bundled_openai_model(GPT_5_6_SOL_OPENAI_MODEL_ID),
AMAZON_BEDROCK_GPT_5_6_SOL_MODEL_ID,
"GPT-5.6 Sol",
/*priority*/ 5,
/*priority*/ 7,
),
bedrock_model(
bundled_openai_model(GPT_5_6_TERRA_OPENAI_MODEL_ID),
AMAZON_BEDROCK_GPT_5_6_TERRA_MODEL_ID,
"GPT-5.6 Terra",
/*priority*/ 6,
/*priority*/ 8,
),
bedrock_model(
bundled_openai_model(GPT_5_6_LUNA_OPENAI_MODEL_ID),
AMAZON_BEDROCK_GPT_5_6_LUNA_MODEL_ID,
"GPT-5.6 Luna",
/*priority*/ 7,
/*priority*/ 9,
),
gpt_5_bedrock_model(
GPT_5_5_OPENAI_MODEL_ID,
AMAZON_BEDROCK_GPT_5_5_MODEL_ID,
"GPT-5.5",
/*priority*/ 8,
/*priority*/ 10,
),
],
};
@@ -112,13 +128,17 @@ pub(super) fn normalize_bundled_bedrock_catalog(
(endpoint, model.slug.as_str()),
(
BedrockEndpoint::Mantle,
AMAZON_BEDROCK_GPT_6_ASTRA_MODEL_ID | AMAZON_BEDROCK_GPT_6_1_SOL_MODEL_ID
AMAZON_BEDROCK_GPT_6_ASTRA_MODEL_ID
| AMAZON_BEDROCK_GPT_6_1_SOL_MODEL_ID
| AMAZON_BEDROCK_GPT_6_2_ASTRA_MODEL_ID
) | (
BedrockEndpoint::Runtime,
"global.openai.gpt-6-astra"
| "us.openai.gpt-6-astra"
| "global.openai.gpt-6.1-sol"
| "us.openai.gpt-6.1-sol"
| "global.openai.gpt-6.2-astra"
| "us.openai.gpt-6.2-astra"
)
) {
model.service_tiers.push(ModelServiceTier {
@@ -205,9 +225,11 @@ mod tests {
.map(|model| model.slug.as_str())
.collect::<Vec<_>>(),
vec![
AMAZON_BEDROCK_GPT_6_2_SOL_MODEL_ID,
AMAZON_BEDROCK_GPT_6_2_ASTRA_MODEL_ID,
AMAZON_BEDROCK_GPT_6_1_TERRA_MODEL_ID,
AMAZON_BEDROCK_GPT_6_1_SOL_MODEL_ID,
AMAZON_BEDROCK_GPT_6_ASTRA_MODEL_ID,
AMAZON_BEDROCK_GPT_6_1_TERRA_MODEL_ID,
AMAZON_BEDROCK_GPT_6_SOL_MODEL_ID,
AMAZON_BEDROCK_GPT_6_LUNA_MODEL_ID,
AMAZON_BEDROCK_GPT_5_6_SOL_MODEL_ID,
@@ -235,14 +257,14 @@ mod tests {
.collect::<Vec<_>>(),
vec![
(
AMAZON_BEDROCK_GPT_6_1_SOL_MODEL_ID,
Some(GPT_5_BEDROCK_CONTEXT_WINDOW),
AMAZON_BEDROCK_GPT_6_2_SOL_MODEL_ID,
Some(500_000),
Some(872_000),
WebSearchToolType::Text,
),
(
AMAZON_BEDROCK_GPT_6_ASTRA_MODEL_ID,
Some(GPT_5_BEDROCK_CONTEXT_WINDOW),
AMAZON_BEDROCK_GPT_6_2_ASTRA_MODEL_ID,
Some(272_000),
Some(872_000),
WebSearchToolType::Text,
),
@@ -252,6 +274,18 @@ mod tests {
Some(872_000),
WebSearchToolType::Text,
),
(
AMAZON_BEDROCK_GPT_6_1_SOL_MODEL_ID,
Some(500_000),
Some(872_000),
WebSearchToolType::Text,
),
(
AMAZON_BEDROCK_GPT_6_ASTRA_MODEL_ID,
Some(GPT_5_BEDROCK_CONTEXT_WINDOW),
Some(872_000),
WebSearchToolType::Text,
),
(
AMAZON_BEDROCK_GPT_6_SOL_MODEL_ID,
Some(GPT_5_BEDROCK_CONTEXT_WINDOW),
@@ -341,11 +375,23 @@ mod tests {
let catalog = static_model_catalog();
for (mut expected, slug, display_name, priority) in [
(
bundled_openai_model(GPT_6_2_SOL_OPENAI_MODEL_ID),
AMAZON_BEDROCK_GPT_6_2_SOL_MODEL_ID,
"GPT-6.2 Sol",
0,
),
(
bundled_openai_model(GPT_6_2_ASTRA_OPENAI_MODEL_ID),
AMAZON_BEDROCK_GPT_6_2_ASTRA_MODEL_ID,
"GPT-6.2 Astra",
1,
),
(
bundled_openai_model(GPT_6_1_SOL_OPENAI_MODEL_ID),
AMAZON_BEDROCK_GPT_6_1_SOL_MODEL_ID,
"GPT-6.1 Sol",
0,
3,
),
(
bundled_openai_model(GPT_6_1_TERRA_OPENAI_MODEL_ID),
@@ -357,37 +403,37 @@ mod tests {
bundled_openai_model(GPT_6_SOL_OPENAI_MODEL_ID),
AMAZON_BEDROCK_GPT_6_SOL_MODEL_ID,
"GPT-6 Sol",
3,
5,
),
(
bundled_openai_model(GPT_6_LUNA_OPENAI_MODEL_ID),
AMAZON_BEDROCK_GPT_6_LUNA_MODEL_ID,
"GPT-6 Luna",
4,
6,
),
(
bundled_openai_model(GPT_5_6_SOL_OPENAI_MODEL_ID),
AMAZON_BEDROCK_GPT_5_6_SOL_MODEL_ID,
"GPT-5.6 Sol",
5,
7,
),
(
bundled_openai_model(GPT_5_6_TERRA_OPENAI_MODEL_ID),
AMAZON_BEDROCK_GPT_5_6_TERRA_MODEL_ID,
"GPT-5.6 Terra",
6,
8,
),
(
bundled_openai_model(GPT_5_6_LUNA_OPENAI_MODEL_ID),
AMAZON_BEDROCK_GPT_5_6_LUNA_MODEL_ID,
"GPT-5.6 Luna",
7,
9,
),
(
bundled_openai_model(GPT_6_ASTRA_OPENAI_MODEL_ID),
AMAZON_BEDROCK_GPT_6_ASTRA_MODEL_ID,
"GPT-6-Astra",
1,
4,
),
] {
expected.slug = slug.to_string();
@@ -401,7 +447,9 @@ mod tests {
expected.additional_speed_tiers.clear();
expected.service_tiers = if matches!(
slug,
AMAZON_BEDROCK_GPT_6_ASTRA_MODEL_ID | AMAZON_BEDROCK_GPT_6_1_SOL_MODEL_ID
AMAZON_BEDROCK_GPT_6_ASTRA_MODEL_ID
| AMAZON_BEDROCK_GPT_6_1_SOL_MODEL_ID
| AMAZON_BEDROCK_GPT_6_2_ASTRA_MODEL_ID
) {
vec![ModelServiceTier {
id: "ultrafast".to_string(),
@@ -3,6 +3,8 @@ use codex_model_provider_info::AMAZON_BEDROCK_GPT_5_6_SOL_MODEL_ID;
use codex_model_provider_info::AMAZON_BEDROCK_GPT_5_6_TERRA_MODEL_ID;
use codex_model_provider_info::AMAZON_BEDROCK_GPT_6_1_SOL_MODEL_ID;
use codex_model_provider_info::AMAZON_BEDROCK_GPT_6_1_TERRA_MODEL_ID;
use codex_model_provider_info::AMAZON_BEDROCK_GPT_6_2_ASTRA_MODEL_ID;
use codex_model_provider_info::AMAZON_BEDROCK_GPT_6_2_SOL_MODEL_ID;
use codex_model_provider_info::AMAZON_BEDROCK_GPT_6_ASTRA_MODEL_ID;
use codex_model_provider_info::AMAZON_BEDROCK_GPT_6_LUNA_MODEL_ID;
use codex_model_provider_info::AMAZON_BEDROCK_GPT_6_SOL_MODEL_ID;
@@ -27,6 +29,8 @@ pub(super) fn static_runtime_model_catalog() -> ModelsResponse {
| AMAZON_BEDROCK_GPT_5_6_LUNA_MODEL_ID
| AMAZON_BEDROCK_GPT_6_1_SOL_MODEL_ID
| AMAZON_BEDROCK_GPT_6_1_TERRA_MODEL_ID
| AMAZON_BEDROCK_GPT_6_2_ASTRA_MODEL_ID
| AMAZON_BEDROCK_GPT_6_2_SOL_MODEL_ID
| AMAZON_BEDROCK_GPT_6_ASTRA_MODEL_ID
| AMAZON_BEDROCK_GPT_6_SOL_MODEL_ID
| AMAZON_BEDROCK_GPT_6_LUNA_MODEL_ID
@@ -18,33 +18,41 @@ fn runtime_catalog_includes_supported_cross_region_models_in_priority_order() {
))
.collect::<Vec<_>>(),
vec![
("global.openai.gpt-6.1-sol", "GPT-6.1 Sol (Global)", 0),
("global.openai.gpt-6-astra", "GPT-6-Astra (Global)", 1),
("global.openai.gpt-6.2-sol", "GPT-6.2 Sol (Global)", 0),
("global.openai.gpt-6.2-astra", "GPT-6.2 Astra (Global)", 1),
("global.openai.gpt-6.1-terra", "GPT-6.1 Terra (Global)", 2),
("global.openai.gpt-6-sol", "GPT-6 Sol (Global)", 3),
("global.openai.gpt-6-luna", "GPT-6 Luna (Global)", 4),
("global.openai.gpt-5.6-sol", "GPT-5.6 Sol (Global)", 5),
("global.openai.gpt-5.6-terra", "GPT-5.6 Terra (Global)", 6),
("global.openai.gpt-5.6-luna", "GPT-5.6 Luna (Global)", 7),
("us.openai.gpt-6.1-sol", "GPT-6.1 Sol (US cross-region)", 8),
("us.openai.gpt-6-astra", "GPT-6-Astra (US cross-region)", 9),
("global.openai.gpt-6.1-sol", "GPT-6.1 Sol (Global)", 3),
("global.openai.gpt-6-astra", "GPT-6-Astra (Global)", 4),
("global.openai.gpt-6-sol", "GPT-6 Sol (Global)", 5),
("global.openai.gpt-6-luna", "GPT-6 Luna (Global)", 6),
("global.openai.gpt-5.6-sol", "GPT-5.6 Sol (Global)", 7),
("global.openai.gpt-5.6-terra", "GPT-5.6 Terra (Global)", 8),
("global.openai.gpt-5.6-luna", "GPT-5.6 Luna (Global)", 9),
("us.openai.gpt-6.2-sol", "GPT-6.2 Sol (US cross-region)", 10),
(
"us.openai.gpt-6.2-astra",
"GPT-6.2 Astra (US cross-region)",
11
),
(
"us.openai.gpt-6.1-terra",
"GPT-6.1 Terra (US cross-region)",
10
12
),
("us.openai.gpt-6-sol", "GPT-6 Sol (US cross-region)", 11),
("us.openai.gpt-6-luna", "GPT-6 Luna (US cross-region)", 12),
("us.openai.gpt-5.6-sol", "GPT-5.6 Sol (US cross-region)", 13),
("us.openai.gpt-6.1-sol", "GPT-6.1 Sol (US cross-region)", 13),
("us.openai.gpt-6-astra", "GPT-6-Astra (US cross-region)", 14),
("us.openai.gpt-6-sol", "GPT-6 Sol (US cross-region)", 15),
("us.openai.gpt-6-luna", "GPT-6 Luna (US cross-region)", 16),
("us.openai.gpt-5.6-sol", "GPT-5.6 Sol (US cross-region)", 17),
(
"us.openai.gpt-5.6-terra",
"GPT-5.6 Terra (US cross-region)",
14
18
),
(
"us.openai.gpt-5.6-luna",
"GPT-5.6 Luna (US cross-region)",
15
19
),
]
);
@@ -66,6 +74,24 @@ fn runtime_catalog_disables_web_search_without_overriding_review_models() {
))
.collect::<Vec<_>>(),
vec![
(
"global.openai.gpt-6.2-sol",
None,
false,
Some(MultiAgentVersion::V2),
),
(
"global.openai.gpt-6.2-astra",
None,
false,
Some(MultiAgentVersion::V2),
),
(
"global.openai.gpt-6.1-terra",
None,
false,
Some(MultiAgentVersion::V2),
),
(
"global.openai.gpt-6.1-sol",
None,
@@ -78,12 +104,6 @@ fn runtime_catalog_disables_web_search_without_overriding_review_models() {
false,
Some(MultiAgentVersion::V2),
),
(
"global.openai.gpt-6.1-terra",
None,
false,
Some(MultiAgentVersion::V2),
),
(
"global.openai.gpt-6-sol",
None,
@@ -114,6 +134,24 @@ fn runtime_catalog_disables_web_search_without_overriding_review_models() {
false,
Some(MultiAgentVersion::V1),
),
(
"us.openai.gpt-6.2-sol",
None,
false,
Some(MultiAgentVersion::V2),
),
(
"us.openai.gpt-6.2-astra",
None,
false,
Some(MultiAgentVersion::V2),
),
(
"us.openai.gpt-6.1-terra",
None,
false,
Some(MultiAgentVersion::V2),
),
(
"us.openai.gpt-6.1-sol",
None,
@@ -126,12 +164,6 @@ fn runtime_catalog_disables_web_search_without_overriding_review_models() {
false,
Some(MultiAgentVersion::V2),
),
(
"us.openai.gpt-6.1-terra",
None,
false,
Some(MultiAgentVersion::V2),
),
(
"us.openai.gpt-6-sol",
None,
+20 -11
View File
@@ -249,16 +249,23 @@ async fn bundled_vertex_models_filter_catalog_and_normalize_unsupported_tool_met
let bundled_models = codex_models_manager::bundled_models_response()
.expect("bundled model catalog should deserialize")
.models;
let mut expected: Vec<_> = ["gpt-6-astra", "gpt-6.1-sol", "gpt-6.1-terra", "gpt-6-luna"]
.into_iter()
.map(|slug| {
bundled_models
.iter()
.find(|model| model.slug == slug)
.expect("bundled catalog should contain the supported Vertex model")
.clone()
})
.collect();
let mut expected: Vec<_> = [
"gpt-6.2-sol",
"gpt-6.2-astra",
"gpt-6-astra",
"gpt-6.1-sol",
"gpt-6.1-terra",
"gpt-6-luna",
]
.into_iter()
.map(|slug| {
bundled_models
.iter()
.find(|model| model.slug == slug)
.expect("bundled catalog should contain the supported Vertex model")
.clone()
})
.collect();
for model in &mut expected {
model.use_responses_lite = false;
model.tool_mode = Some(ToolMode::Direct);
@@ -274,10 +281,12 @@ async fn bundled_vertex_models_filter_catalog_and_normalize_unsupported_tool_met
.map(|model| (model.model, model.show_in_picker))
.collect::<Vec<_>>(),
vec![
("gpt-6.2-sol".to_string(), true),
("gpt-6.2-astra".to_string(), true),
("gpt-6.1-terra".to_string(), true),
("gpt-6.1-sol".to_string(), true),
("gpt-6-astra".to_string(), true),
("gpt-6-luna".to_string(), true),
("gpt-6.1-terra".to_string(), true),
]
);
}
+6 -1
View File
@@ -160,7 +160,12 @@ impl ModelProvider for GcpModelProvider {
model_catalog.models.retain(|model| {
matches!(
model.slug.as_str(),
"gpt-6-astra" | "gpt-6-luna" | "gpt-6.1-sol" | "gpt-6.1-terra"
"gpt-6-astra"
| "gpt-6-luna"
| "gpt-6.1-sol"
| "gpt-6.1-terra"
| "gpt-6.2-astra"
| "gpt-6.2-sol"
)
});
// Vertex supports function tools, namespaces, and client-side tool search, but not
@@ -761,6 +761,8 @@ async fn amazon_bedrock_provider_creates_static_models_manager() {
assert_eq!(uncached_catalog, catalog);
for slug in [
"openai.gpt-6.1-sol",
"openai.gpt-6.2-astra",
"openai.gpt-6.2-sol",
"openai.gpt-6.1-terra",
"openai.gpt-6-sol",
"openai.gpt-6-luna",
@@ -792,9 +794,11 @@ async fn amazon_bedrock_provider_creates_static_models_manager() {
assert_eq!(
models,
vec![
("openai.gpt-6.2-sol", "GPT-6.2 Sol"),
("openai.gpt-6.2-astra", "GPT-6.2 Astra"),
("openai.gpt-6.1-terra", "GPT-6.1 Terra"),
("openai.gpt-6.1-sol", "GPT-6.1 Sol"),
("openai.gpt-6-astra", "GPT-6-Astra"),
("openai.gpt-6.1-terra", "GPT-6.1 Terra"),
("openai.gpt-6-sol", "GPT-6 Sol"),
("openai.gpt-6-luna", "GPT-6 Luna"),
("openai.gpt-5.6-sol", "GPT-5.6 Sol"),
@@ -816,9 +820,11 @@ async fn amazon_bedrock_provider_creates_static_models_manager() {
.map(|preset| preset.model.as_str())
.collect::<Vec<_>>(),
vec![
"openai.gpt-6.2-sol",
"openai.gpt-6.2-astra",
"openai.gpt-6.1-terra",
"openai.gpt-6.1-sol",
"openai.gpt-6-astra",
"openai.gpt-6.1-terra",
"openai.gpt-6-sol",
"openai.gpt-6-luna",
"openai.gpt-5.6-sol",
@@ -833,7 +839,7 @@ async fn amazon_bedrock_provider_creates_static_models_manager() {
.find(|preset| preset.is_default)
.expect("Bedrock catalog should have a default model");
assert_eq!(default_model.model, "openai.gpt-6.1-sol");
assert_eq!(default_model.model, "openai.gpt-6.2-sol");
}
#[tokio::test]
File diff suppressed because one or more lines are too long
@@ -14,7 +14,7 @@ attached:
› draft during startup
GPT-6.1-Sol default · /tmp/project
GPT-6.2-Sol default · /tmp/project
owned=true
startup:
@@ -28,4 +28,4 @@ attached:
› draft during startup
GPT-6.1-Sol default · /tmp/project
GPT-6.2-Sol default · /tmp/project
@@ -733,7 +733,7 @@ async fn fresh_startup_reads_destination_and_cleared_model_uses_catalog() -> Res
insta::assert_snapshot!(rendered, @"
› Ask Codex to do anything
GPT-6.1-Sol high · <PROJECT>
GPT-6.2-Sol high · <PROJECT>
? for shortcuts
");
}
@@ -5,7 +5,7 @@ expression: "render_bottom_popup(&chat, 100)"
Select Model and Effort
› 1. GPT-5.6 Terra (current) Older balanced model for straightforward work.
2. GPT-5.6 Luna Older fast and efficient model.
› 1. GPT-5.6 Terra (current) Legacy balanced model for agentic coding.
2. GPT-5.6 Luna Legacy fast model for simpler tasks.
enter select · esc back
@@ -0,0 +1,17 @@
---
source: tui/src/chatwidget/tests/bedrock_catalog_tests.rs
expression: "format!(\"Bundled catalog:\\n{bundled_view}\\n\\nCustom catalog:\\n{custom_view}\")"
---
Bundled catalog:
› /ultrafast The fastest available responses for latency-sensitive work.
› /ultrafast
GPT-6.2 Astra default · /tmp/project
Custom catalog:
› /express My custom Bedrock tier.
› /express
GPT-6.2-Astra default · /tmp/project
@@ -5,13 +5,13 @@ expression: "render_bottom_popup(&chat, 100)"
Select Model and Effort
› 1. GPT-6.1 Sol (current) Latest workhorse model for coding and everyday work.
2. GPT-6-Astra Frontier intelligence for the most demanding work.
3. GPT-6.1 Terra GPT-6.1 Terra Codex model.
4. GPT-6 Sol Previous generation workhorse model.
5. GPT-6 Luna Fast and affordable model for easier tasks.
6. GPT-5.6 Sol Older generation workhorse model.
7. GPT-5.6 Terra Older balanced model for straightforward work.
8. GPT-5.6 Luna Older fast and efficient model.
› 1. GPT-6.2 Sol (current) Fast, go-to model for coding and everyday work.
2. GPT-6.2 Astra Our most capable model for complex coding and professional work.
3. GPT-6.1 Terra Efficient model that excels at visual understanding and computer use.
4. GPT-6.1 Sol Previous generation model for coding and everyday work.
5. GPT-6-Astra Previous generation model for complex coding and professional work.
6. GPT-6 Sol Legacy model for coding and everyday work.
7. GPT-6 Luna Fast and affordable model for easier tasks.
8. GPT-5.6 Sol Legacy model for coding and everyday work.
↓
enter select · esc back
@@ -0,0 +1,17 @@
---
source: tui/src/chatwidget/tests/bedrock_catalog_tests.rs
expression: "format!(\"Bundled catalog:\\n{bundled_view}\\n\\nCustom catalog:\\n{custom_view}\")"
---
Bundled catalog:
› /ultrafast The fastest available responses for latency-sensitive work.
› /ultrafast
GPT-6.2 Astra (Global) default · /tmp/project
Custom catalog:
› /express My custom Bedrock tier.
› /express
GPT-6.2-Astra default · /tmp/project
@@ -5,13 +5,16 @@ expression: "render_bottom_popup(&chat, 100)"
Select Model and Effort
› 1. GPT-6.1 Sol (Global) (current) Latest workhorse model for coding and everyday work.
2. GPT-6-Astra (Global) Frontier intelligence for the most demanding work.
3. GPT-6.1 Terra (Global) GPT-6.1 Terra Codex model.
4. GPT-6 Sol (Global) Previous generation workhorse model.
5. GPT-6 Luna (Global) Fast and affordable model for easier tasks.
6. GPT-5.6 Sol (Global) Older generation workhorse model.
7. GPT-5.6 Terra (Global) Older balanced model for straightforward work.
8. GPT-5.6 Luna (Global) Older fast and efficient model.
› 1. GPT-6.2 Sol (Global) (current) Fast, go-to model for coding and everyday work.
2. GPT-6.2 Astra (Global) Our most capable model for complex coding and professional
work.
3. GPT-6.1 Terra (Global) Efficient model that excels at visual understanding and
computer use.
4. GPT-6.1 Sol (Global) Previous generation model for coding and everyday work.
5. GPT-6-Astra (Global) Previous generation model for complex coding and professional
work.
6. GPT-6 Sol (Global) Legacy model for coding and everyday work.
7. GPT-6 Luna (Global) Fast and affordable model for easier tasks.
8. GPT-5.6 Sol (Global) Legacy model for coding and everyday work.
↓
enter select · esc back
@@ -0,0 +1,17 @@
---
source: tui/src/chatwidget/tests/bedrock_catalog_tests.rs
expression: "format!(\"Bundled catalog:\\n{bundled_view}\\n\\nCustom catalog:\\n{custom_view}\")"
---
Bundled catalog:
› /ultrafast The fastest available responses for latency-sensitive work.
› /ultrafast
GPT-6.2 Astra (US cross-region) default · /tmp/project
Custom catalog:
› /express My custom Bedrock tier.
› /express
GPT-6.2-Astra default · /tmp/project
@@ -10,5 +10,5 @@ expression: "format!(\"{}\\n{}\", lines_to_single_string(cells.last().unwrap()),
› Viewing sub-agent — direct input is disabled
GPT-6.1-Sol default · /tmp/project
GPT-6.2-Sol default · /tmp/project
? for shortcuts
@@ -5,8 +5,7 @@ expression: "render_bottom_popup(&chat, 80)"
Select Model and Effort
› 1. Renamed model Older balanced model for straightforward
work.
› 1. Renamed model Legacy balanced model for agentic coding.
2. Shared display name (current) Legacy coding model.
enter select · esc back
@@ -6,13 +6,14 @@ expression: popup
Select Model and Effort
↑
2. GPT-6-Astra Frontier intelligence for the most demanding work.
3. GPT-6-Sol Previous generation workhorse model.
4. GPT-6-Luna Fast and affordable model for easier tasks.
5. GPT-5.6-Sol Older generation workhorse model.
6. GPT-6.1 Terra GPT-6.1 Terra Codex model.
7. GPT-5.6-Terra Older balanced model for straightforward work.
8. GPT-5.6-Luna Older fast and efficient model.
› 9. GPT-5.5 (current) Legacy coding model.
4. GPT-6.1-Sol Previous generation model for coding and everyday work.
5. GPT-6-Astra Previous generation model for complex coding and
professional work.
6. GPT-6-Sol Legacy model for coding and everyday work.
7. GPT-6-Luna Fast and affordable model for easier tasks.
8. GPT-5.6-Sol Legacy model for coding and everyday work.
9. GPT-5.6-Terra Legacy balanced model for agentic coding.
GPT-5.6-Luna Legacy fast model for simpler tasks.
› GPT-5.5 (current) Legacy coding model.
enter select · esc back
@@ -13,13 +13,13 @@ async fn bedrock_astra_model_and_reasoning_pickers() {
(
"mantle",
ModelProviderInfo::create_amazon_bedrock_provider(/*aws*/ None),
"openai.gpt-6.1-sol",
"openai.gpt-6.2-sol",
"openai.gpt-6-astra",
),
(
"runtime",
ModelProviderInfo::create_amazon_bedrock_runtime_provider(/*aws*/ None),
"global.openai.gpt-6.1-sol",
"global.openai.gpt-6.2-sol",
"global.openai.gpt-6-astra",
),
] {
@@ -157,6 +157,24 @@ async fn bedrock_ultrafast_slash_command_selects_and_clears_tier() {
"us.openai.gpt-6.1-sol",
"us.openai.gpt-6.1-terra",
),
(
"mantle_astra62",
ModelProviderInfo::create_amazon_bedrock_provider(/*aws*/ None),
"openai.gpt-6.2-astra",
"openai.gpt-6.2-sol",
),
(
"runtime_global_astra62",
ModelProviderInfo::create_amazon_bedrock_runtime_provider(/*aws*/ None),
"global.openai.gpt-6.2-astra",
"global.openai.gpt-6.2-sol",
),
(
"runtime_us_astra62",
ModelProviderInfo::create_amazon_bedrock_runtime_provider(/*aws*/ None),
"us.openai.gpt-6.2-astra",
"us.openai.gpt-6.2-sol",
),
] {
let presets = create_model_provider(provider_info.clone(), /*auth_manager*/ None)
.models_manager_without_cache(/*config_model_catalog*/ None)
@@ -10,7 +10,7 @@ expression: "render_bottom_popup(&app.chat_widget,\n96).replace(&format!(\"{proj
› ● Build the dashboard │ Build the dashboard
● Repair authentication │ ● Needs input
○ Review yesterday's changes │
│ Model: gpt-6.1-sol
│ Model: gpt-6.2-sol
│ Reasoning: Unknown
│
│ Project
@@ -2,4 +2,4 @@
source: tui/src/app/tests.rs
expression: "frame.lines().find(|row| row.contains(\"Find:\")).unwrap()"
---
Find: parent-2x4 │ GPT-6.1-Sol default
Find: parent-2x4 │ GPT-6.2-Sol default
@@ -19,7 +19,7 @@ parent-29
› parent draft
GPT-6.1-Sol default · Main [default]
GPT-6.2-Sol default · Main [default]
New activity · esc latest
100 columns, reopened
@@ -39,8 +39,8 @@ parent-29 │
│
› parent draft- │
│› child draft!
GPT-6.1-Sol default │
│ GPT-6.1-Sol default · Agent
GPT-6.2-Sol default │
│ GPT-6.2-Sol default · Agent
99 columns, reopened
Side chat × +
@@ -59,5 +59,5 @@ child-29
› child draft!
GPT-6.1-Sol default · Agent
GPT-6.2-Sol default · Agent
Side from main thread · ⌃/ to switch · ⌃c to close
@@ -19,7 +19,7 @@ Zoomed
› xchild draft!
GPT-6.1-Sol default · Agent
GPT-6.2-Sol default · Agent
tab to queue message · Side from main thread · ⌃/ to switch · ⌃c to close
Restored
parent-24 │ Side chat × +
@@ -38,7 +38,7 @@ parent-29
│
› xparent draft-! │› xchild draft!
second line │
GPT-6.1-Sol default │ GPT-6.1-Sol default · Agent
GPT-6.2-Sol default │ GPT-6.2-Sol default · Agent
│ tab to queue message · Side from main thread · ⌃/ to switch · ⌃c to close
Equalized
parent-24 │ Side chat × +
@@ -57,5 +57,5 @@ parent-29
│
› xparent draft-! │› xchild draft!
second line │
GPT-6.1-Sol default │ GPT-6.1-Sol default · Agent
GPT-6.2-Sol default │ GPT-6.2-Sol default · Agent
│ tab to queue message · Side from main thread · ⌃/ to switch · ⌃c to close
@@ -5,5 +5,5 @@ expression: "render_bottom_popup(&app.chat_widget, 80)"
---
› Ask Codex to do anything
GPT-6.1-Sol default · /tmp/project · child-0 [worker]
GPT-6.2-Sol default · /tmp/project · child-0 [worker]
← for agents · ? for shortcuts
@@ -5,5 +5,5 @@ expression: "render_bottom_popup(&app.chat_widget, 80)"
---
› Viewing sub-agent — direct input is disabled
GPT-6.1-Sol default · /tmp/project · child-1 [worker]
GPT-6.2-Sol default · /tmp/project · child-1 [worker]
← for agents · ? for shortcuts
@@ -18,4 +18,4 @@ parent-29 │
│
› parent draft │› child draft!
│
GPT-6.1-Sol default · Main [default] │ GPT-6.1-Sol default
GPT-6.2-Sol default · Main [default] │ GPT-6.2-Sol default