## What changed
- Add optional `default` and `plan` collaboration-mode messages to model catalog metadata.
- Prefer the active mode's catalog message, including an explicit empty value, while falling back to the existing developer instructions when the catalog variant is absent.
- Track the model in collaboration-mode world state so mode and model changes append the appropriate instructions, and clear prior instructions when no current message exists.
- Preserve catalog collaboration messages when instruction overrides remove other model instruction fields.
## Testing
- Cover catalog deserialization, variant selection, fallback and empty-message behavior, legacy snapshot migration, mode changes, and model changes.
GitOrigin-RevId: d4c6c0028423bc50b0e8175e55586f3e7bc1c3a8
## What changed
- Preserve `encrypted_function_args` on function calls so an empty list can mark plaintext collaboration arguments across request replay.
- Deliver `spawn_agent`, `send_message`, and `followup_task` payloads as structured plaintext agent messages when that marker is present; retain encrypted delivery otherwise.
- Redact plaintext collaboration arguments from tool and communication logs, and omit the metadata when sending requests to non-OpenAI providers.
## Testing
- Cover serialization of empty encrypted-argument metadata and plaintext versus encrypted subagent message delivery.
- Verify plaintext tool arguments are redacted and provider-specific metadata is removed from non-OpenAI requests.
GitOrigin-RevId: 64db98ff0b61a3af2f04ed609292363f2e2362a8
## What changed
- Propagate the initiating turn ID through agent spawns, follow-up tasks, reviews, and delegated Codex sessions.
- Add `parent_turn_id` to Responses client and turn metadata while keeping it out of external MCP metadata.
- Preserve parent-turn provenance across queued agent messages when their triggering parent is unambiguous.
## Testing
- Cover spawned, resumed, nested, reviewed, delegated, and WebSocket request metadata.
- Verify queued messages do not claim ambiguous or queue-only parent turns.
GitOrigin-RevId: 481fdebbe7df2031880fe259509273cce50b20a8
## What changed
- Add token-budget settings to model catalog messages and apply them when the
feature is enabled without explicit token-budget configuration.
- Keep explicit user settings authoritative, reject invalid catalog defaults,
and preserve resolved defaults in exported config locks.
- Manage context-window guidance through world state so it updates once when
the active model changes while retaining prior conversation history.
## Testing
- Cover catalog defaults, explicit overrides, disabled features, invalid
values, config-lock replay, and model switching.
GitOrigin-RevId: 54544fefaa14b09f5ef5bad9967a13c52b87b0fe
## What changed
- Scope pending network approvals to a turn and execution while still coalescing duplicate requests within one execution.
- Fail abandoned approval owners closed, cancel denied executions, resolve any waiting requests, and cancel in-flight Guardian reviews when their owner is dropped.
- Serialize session policy updates with approval caches so concurrent allow and deny decisions cannot leave enforcement and cached decisions inconsistent.
- Give user approval requests and permission hooks unique per-execution identifiers.
## Testing
- Cover approval deduplication boundaries, abandoned-owner cleanup, cancellation outcomes, replacement requests, and Guardian cancellation events.
GitOrigin-RevId: 11e184dd1c70294a99581b6ff0d4b6607d9ea1d9
## Why
MCP elicitation reviewers are reused across runtime refreshes, while an active
turn can retain the approval settings it started with. Reviewing against that
turn could therefore apply stale authority after session settings changed.
## What changed
- Read the latest published MCP runtime configuration when reviewing an
elicitation, including the approval policy, permission profile, configuration
layers, and reviewer selection.
- Apply `never` and granular MCP-elicitation policy decisions before routing an
eligible request to Guardian.
- Keep MCP runtime startup registered for refresh invalidation through session
creation.
## Testing
Added coverage that refreshes a running session and verifies the same reviewer
uses the latest authority for Guardian routing, denial, and empty-form
auto-approval.
GitOrigin-RevId: 826a157aa70fc8e3a9e3a14ecb0261bf7b63f63a
## Why
Guardian review commands can run without the proxy-port environment used to
configure the parent session. Reconciling persistent Windows sandbox settings
for those commands can discard the parent's proxy configuration.
## What changed
- Add a session-level Windows sandbox proxy-settings mode and use `Preserve`
for guardian review sessions while keeping `Reconcile` as the default.
- Carry the mode through unified exec and the exec-server sandbox context to
Windows process launches.
- In preserve mode, reuse the proxy settings recorded by the existing sandbox
setup marker.
## Testing
Added coverage for guardian session configuration, exec-server transport, and
preserving an existing setup marker when proxy-port environment variables are
absent.
GitOrigin-RevId: 214655c6e6d97360906431773dc0de0fdda2db6e
## What changed
- Add optional `plugin_id` and `script_path` fields to execution approval and guardian assessment events.
- Propagate validated plugin attribution through delegated approvals, guardian-reviewed command items, app-server notifications, thread history, and rollout traces.
- Preserve attribution on both started and completed command items, including declined commands.
## Testing
- Extend core, app-server, thread-history, and rollout-trace tests to cover plugin attribution propagation and serialization.
GitOrigin-RevId: 723684d010cab04142918d7a95e06ed95d008da1
## Why
Parent context-window and auto-compaction overrides may describe a different
model than the one selected for Guardian review.
## What changed
- Clear `model_context_window` and `model_auto_compact_token_limit` when the
effective Guardian model differs from the parent model, allowing the review
model's own limits to apply.
- Preserve those overrides when Guardian uses the same effective model.
## Testing
Add unit coverage for both model-selection paths and verify that a Guardian
rollout records the selected review model's context window.
GitOrigin-RevId: ebc399f4eba37c7f279fd0d258154c00d45e2d00
## What changed
- Store the spawned session's working directory as a `PathUri` in
`GuardianReviewSessionReuseKey` so reuse comparisons use its canonical URI
representation.
- Assert that the reuse key contains the converted working directory.
GitOrigin-RevId: 68d440a37ebb310c610dabbb3373ed0ceff88dfa
## What changed
- Remove the step-scoped `ExtensionData` argument from context, turn-input, and tool contributors.
- Pass the host's optional `McpResourceClient` through `ThreadStartInput` so extensions can retain session capabilities explicitly.
- Keep the MCP resource client in skills-owned session state for catalog loading and skill tools.
GitOrigin-RevId: bafa77bcd998aff408d6a396c5fd9ac268c4cce4
## Why
Startup prewarms and Guardian review sessions do not need workspace Git metadata. Collecting it for those requests performs redundant work, while user turns still need a fresh view of the repository.
## What changed
- Add a per-session Git enrichment policy.
- Skip enrichment for startup prewarm, preview, and Guardian review turns.
- Keep fresh enrichment enabled for regular sessions and one-shot delegated turns.
## Testing
- Cover prewarm and Guardian requests omitting workspace metadata while subsequent user turns observe repository changes.
- Cover concurrent repositories and worktrees retaining distinct metadata.
- Verify concurrent enrichment is coalesced and canceled enrichment can be retried.
GitOrigin-RevId: 4a68615288bb568c75f2e487d9c0776e3e80f191
## Why
Default read-only protections for project metadata should apply when paths such
as `.git`, `.agents`, and `.codex` exist, without causing sandbox setup to
materialize missing paths as ACL targets.
## What changed
- Add an optional `missing_path_behavior` to filesystem sandbox entries and
mark default project-metadata protections with `skip`.
- Preserve the behavior through permission transforms and exec/MCP protocol
serialization while keeping existing path wire variants stable.
- Ignore skip-missing entries when projecting configuration or Windows sandbox
overrides, while retaining explicit metadata carveouts.
## Testing
- Cover protocol round trips for path and special-path entries.
- Verify default metadata protections and Windows explicit carveout handling.
GitOrigin-RevId: 6df13dadacdd131c44aab9f15a967c81051355c1
## What changed
- Add an `ExtensionData` store to each `StepContext`.
- Pass the step store to context, world-state, turn-input, and tool contributors so extensions can use capabilities bound to the current sampling step.
- Preserve the captured step context when rebuilding initial context during compaction.
GitOrigin-RevId: 51f3105d3a793f60534f122f9be8247c131c2ca5
## What changed
- Remove the `spawn_agents_on_csv` and `report_agent_job_result` tools and their agent-job runtime and state models.
- Drop the legacy `agent_jobs` and `agent_job_items` tables during state database migration.
- Keep `features.enable_fanout` and `agents.job_max_runtime_seconds` accepted as no-op compatibility settings while omitting them from the generated configuration schema.
## Testing
- Verify upgrades remove both legacy agent-job tables.
- Verify the removed feature and configuration keys still parse without taking effect.
GitOrigin-RevId: 8cc3337da78c67162229f02f40a747f503542646
## What changed
- Change `ReviewDecision::Denied` to carry a rejection string and update the generated protocol schemas.
- Preserve specific rejection reasons through command, patch, network, MCP, delegated, and automatic approval flows so tool results can return them to the model.
- Distinguish invalid approval responses from user declines and truncate oversized rejection messages before adding them to model context.
## Testing
- Add coverage for denial serialization, invalid app-server approval responses, dropped automatic reviews, network rejection propagation, and rejection-message truncation.
GitOrigin-RevId: d48a55643aa03c7ecfa08a8e4b669913eec5464a
## What changed
- Add remote and local audio variants to core and app-server user inputs, including generated JSON and TypeScript schemas.
- Preserve audio fields when converting app-server inputs and response content into thread items.
- Keep unsupported audio out of model requests, TUI rendering, text extraction, and truncated tool output, using explicit text placeholders where audio is submitted as a model input.
## Testing
- Cover app-server media conversion, unsupported-input placeholders, TUI omission, and tool-output truncation.
GitOrigin-RevId: 16152bf3af674575ce569764a8b472e032e7569d
## What changed
- Parse `cache_write_tokens` from response input token details and carry it through token usage aggregation.
- Expose `cache_write_input_tokens` in protocol, app-server, exec, and TypeScript SDK usage events, defaulting it to zero for compatibility with older payloads.
- Report cache-write usage in analytics, metrics, traces, and rollout data alongside existing cache-read usage.
## Testing
- Add coverage for response parsing, usage event output, aggregation, analytics, metrics, and telemetry fields.
GitOrigin-RevId: 634e4fbe3086a8923c0ee13b622573e7d9d221bd
## What changed
- Replace the combined `Codex` wrapper with an `Arc<Session>` and a dedicated
`SessionIo` handle for submissions, events, status, and loop termination.
- Make `Session::spawn` return the session and I/O handle separately, and have
`CodexThread` own both.
- Update thread management, delegated sessions, guardian reviews, and agent
control to use the appropriate session or I/O interface.
GitOrigin-RevId: 8523bad7e7366e2969fd0cc3849b4247b141427a
## What changed
- Add an optional `policy_template` field to auto-review model messages.
- Build Guardian instructions from the catalog template when present, while
preserving the existing precedence for managed and catalog policy content.
- Fall back independently to the bundled template and bundled policy, and
preserve explicit empty catalog values.
## Testing
- Cover missing and empty template deserialization, policy precedence, bundled
fallbacks, and Guardian prewarm instructions.
GitOrigin-RevId: 93116dadbc9bf40527882844a428caed4159fff6
## What changed
- Add per-sandbox-mode permission messages to `ModelMessages` and preserve them when applying model configuration overrides.
- Use the selected catalog message in permission instructions, substituting `{{ network_access }}` with the active network policy. Fall back to the existing sandbox text when no override is provided, and allow an empty override to omit only the sandbox section.
- Apply catalog permission messages when a session starts and when its model changes.
## Testing
- Cover catalog deserialization, mode selection, network substitution, fallback and empty-message behavior, remote model catalogs, and model changes.
GitOrigin-RevId: cd5ed3aee3155dca0e7b2358c0f09bae73236856
## What changed
- Limit Guardian reviewer sessions with an environment to `exec_command`,
`write_stdin`, and `view_image`.
- Exclude general tool sources from Guardian reviewer tool planning.
## Testing
- Extend the Guardian request snapshot test to assert the exact model-visible
tool list.
GitOrigin-RevId: 80ba7011c23c32d163aaad906087fac396cc70fa
## What changed
- Add an `auto_review.policy` field to model catalog messages.
- Use the selected Guardian model's catalog policy for review-session instructions, while preserving the precedence of `guardian_policy_config` and falling back to the built-in policy when neither is present.
- Preserve auto-review messages when model instruction overrides remove catalog instruction templates.
## Testing
- Cover configured-policy precedence, explicit empty catalog policies, catalog-message preservation, and propagation of the catalog policy into a prewarmed Guardian session.
GitOrigin-RevId: 26b61ae2958ea8325a64834dcf91f47e140d74b3
## What changed
- Include permission instructions in Guardian review requests and let the review model use its configured tool mode and standard tool plan instead of a Guardian-specific direct-tool override.
- Refine the Guardian policy for tenant policy precedence, authorization scoring, prompt-injection handling, read-only investigation, and post-denial user approval.
- Simplify low-risk allow responses to `{"outcome":"allow"}` regardless of authorization scoring.
## Testing
- Update Guardian request-layout snapshots to cover the permission instructions included in initial and follow-up reviews.
GitOrigin-RevId: eb83571ef61f7b4e854c12e13e7163c6fab43f00
## What changed
- Emit the thread-idle extension lifecycle when the guardian successfully aborts an active turn after repeated automatic review denials.
- Keep user-initiated interrupt behavior unchanged.
- Add a regression test that waits for the thread-idle callback after a guardian interrupt.
GitOrigin-RevId: 6622c35361a3da9126c11bd6ca49a532c1bd6e27
## What changed
- Add an optional `error` payload to `TurnCompleteEvent` and omit it when a turn completes without an error.
- Preserve the full terminal `ErrorEvent` through the turn lifecycle so the completion event includes the same error details emitted separately by `EventMsg::Error`.
## Testing
- Extend the stream-error integration test to verify that `TurnCompleteEvent` contains the emitted error and that the next turn can still proceed.
GitOrigin-RevId: 8d32942a44132763076130102d8b89ab356d59d0
## What changed
- Add an optional `started_at` Unix timestamp to `TurnCompleteEvent` and
`TurnAbortedEvent`.
- Populate the timestamp from turn timing state when a turn completes or is
aborted.
- Preserve the timestamp when synthesizing interrupted fork history and when
importing external sessions.
## Testing
- Verify that interrupting a turn emits a `TurnAbortedEvent` with a start
timestamp.
GitOrigin-RevId: 75b3911b95c060e5fc1843a5e293d8fe47377186
## What changed
- Build a reasoning payload for every Responses request and always include `reasoning.encrypted_content`.
- Remove `supports_reasoning_summaries` from model metadata and retire the `model_supports_reasoning_summaries` configuration override.
- Use configured or model-default reasoning effort without a capability gate, including for guardian reviews and tracing.
GitOrigin-RevId: 2c9f194a5d2d4d688a2235299e6358f82ab8e1ea
## Why
`NetworkProxyConfig` only wrapped `NetworkProxySettings` in a single
`network` field. That extra level made runtime callers repeat `.network`
everywhere without representing a real boundary.
## What changed
- move the managed-network fields directly onto `NetworkProxyConfig`
- collapse the matching partial-config wrapper
- update runtime callers and tests to use the direct fields
- keep the user-facing permissions/profile TOML layout unchanged
The internal serialized shape now matches the runtime type itself. This
does not change managed-network behavior or the `config.toml` shape.
## Why
Auto-review performance is weaker because of confusing instructions
about sandbox permissions, and because it is given many tools which are
irrelevant to it.
## What
* Update the auto review prompt
* Remove the permissions_instructions developer message
* Only pass exec_tool and view_image tool to the reviewer
## Validation
`just fmt`
`cargo test -p codex-core --lib --quiet`
## Why
With deferred executors, a turn can begin before a remote environment
attaches. AGENTS.md discovery previously ran only during session setup,
so instructions from a later environment never reached the model or the
session instruction sources.
WorldState persistence has now landed, so this uses the durable
model-visible baseline directly instead of carrying a temporary
resume/fork compatibility path.
## What
- Add an `AgentsMdManager` in `SessionServices` to own host
instructions, loaded state, and refresh caching.
- When `DeferredExecutor` is enabled, refresh AGENTS.md when attached
environment selections change and freeze the result in the corresponding
`StepContext`.
- Represent AGENTS.md as a persisted WorldState section for every
session, with bounded initial, replacement, and removal updates.
- Remove duplicate AGENTS.md state and rendering from
`SessionConfiguration` and `TurnContext`.
- Build initial context, per-request updates, and compaction context
from the same step-scoped value.
- On resume and fork, compare current instructions with the restored
WorldState baseline and inject a replacement exactly once when they
differ.
Builds on #29833, #29835, and #29837.
## Tests
- Covers a remote environment becoming ready mid-turn, with AGENTS.md
appearing on the next request exactly once and updating canonical
instruction sources.
- Covers full, unchanged, replaced, and removed AGENTS.md WorldState
rendering.
- Covers changed instructions across cold resume and fork without
duplicate reinjection.
- Covers remote-v2 compaction retaining creation-time instructions in
the live session and cold resume appending one replacement when the
source changed.
- Ran focused `codex-core` AGENTS.md, WorldState, and context-update
test suites.
## Why
auto review reviews Codex App tool calls using connector metadata such
as the app ID, name, and description. That metadata does not identify
the account behind the OAuth connection.
For Google Drive, this means auto review cannot distinguish a Drive
connection authenticated as `user@email.com` from a personal Drive
account. Uploading work data can therefore look like a transfer to a
personal destination even though the connector service already knows the
authenticated account email.
## What changed
- Read `_meta._codex_apps.connected_account_email` while resolving
approval metadata for built-in Codex App tools.
- Include the connected account email in the structured MCP tool action
sent to auto review.
- Trim empty values and omit the field when the connector link has no
account email.
- Update existing auto review request constructors and add coverage for
request construction and JSON serialization.
## Security
Only metadata from the trusted built-in `codex_apps` MCP server is
accepted. Custom MCP servers cannot inject a connected account email
into auto review reviews; the new regression test verifies that spoofed
metadata is ignored.
The email is used only in auto review's private review request. This
change does not add it to model-visible tool descriptions, app-server
approval events, or auto review assessment/review analytics.
## Why
The first auto-review currently creates its Guardian child session on
demand, adding avoidable latency before the review can begin. Creating
the ordinary Guardian child during parent-session initialization lets
that child use the existing session startup WebSocket prewarm before the
first escalation. This does not introduce a Guardian-specific prewarm
mechanism.
## What changed
- initialize the existing Guardian review-session manager owned by
`Session` when a thread starts with auto-review enabled and an approval
policy that routes to Guardian
- use the standard Guardian child-session construction and the existing
session startup WebSocket prewarm
- preserve the existing reuse-key invalidation and lazy creation
fallback when startup initialization fails or the effective review
configuration changes
- add an integration test that verifies normal root-session startup
emits a Guardian `generate=false` prewarm request
## Benchmark
I compared release builds against main. Each prompt first ran a
non-escalated `sleep 3`, then requested an escalated marker command.
| binary | count | avg Guardian duration | median Guardian duration |
avg Guardian TTFT |
|---|---:|---:|---:|---:|
| origin-main | 10 | 4008.7 ms | 3949.5 ms | 3746.5 ms |
| session-fix | 10 | 2865.0 ms | 2594.0 ms | 2492.7 ms |
Guardian duration fell by 28.5% and Guardian TTFT fell by 33.5%. These
measurements cover Guardian review latency; they do not measure parent
thread-start latency.
## Description
This PR cuts Codex over from generic `ResponseItem.metadata` (introduced
here: https://github.com/openai/codex/pull/28355) to
`ResponseItem.internal_chat_message_metadata_passthrough`, which is the
blessed path and has strongly-typed keys.
For now we have to drop this MAv2 usage of `metadata`:
https://github.com/openai/codex/pull/28561 until we figure out where
that should live.
## Summary
- keep context injection on a single ContextContributor trait
- split context injection into thread-scoped and turn-scoped
contribution methods
- wire turn-scoped fragments into initial context assembly so extensions
can contribute context from turn-local state
## Why
`ResponseItem` variants do not have a consistent internal ID shape: some
variants carry required IDs, some carry optional IDs, and some cannot
represent an ID at all. The existing fields also use inconsistent serde,
TypeScript, and JSON-schema annotations. A single enum-level access path
is needed before history recording can assign and retain IDs.
This PR establishes that internal model only. It intentionally does not
generate or serialize IDs; allocation and wire persistence are isolated
in the stacked follow-up.
## What changed
- Give every concrete `ResponseItem` variant an `Option<String>` ID
field.
- Apply the same internal-only annotations to every ID field:
`#[serde(default, skip_serializing)]`, `#[ts(skip)]`, and
`#[schemars(skip)]`.
- Add `ResponseItem::id()` and `ResponseItem::set_id()` as the shared
accessors.
- Preserve IDs when history items are rewritten for truncation.
- Adapt consumers that previously assumed reasoning and image-generation
IDs were required.
- Regenerate app-server schemas so the hidden fields are represented
consistently.
The serde catch-all `ResponseItem::Other` remains ID-less because it
must remain a unit variant.
## Test plan
- `cargo check --tests -p codex-core -p codex-api -p codex-rollout-trace
-p codex-image-generation-extension`
- `just test -p codex-protocol`
- `just test -p codex-app-server-protocol`
- `just test -p codex-api -p codex-rollout-trace -p
codex-image-generation-extension`
- `just test -p codex-core event_mapping`
## Why
Host skill discovery was still exposed as a manager even though it is a
process-owned service shared by sessions, the app-server catalog, and
file-watcher invalidation. The skills extension also consumed an ad hoc
loaded-skills wrapper instead of a named immutable snapshot.
## What changed
- replace `SkillsManager` with concrete `SkillsService`
- make the service cache and return immutable `HostSkillsSnapshot`
values
- migrate the skills extension host provider to the snapshot boundary
- migrate app-server catalog, watcher, and invalidation paths to the
service
This keeps the service limited to host discovery, caching, roots, and
invalidation. Catalog rendering and invocation remain extension
responsibilities for the next stacked change.
## Why
app-server may run on a different OS from the selected exec-server
environment. Parsing that environment’s cwd with the Codex host’s path
rules prevents thread startup.
## What
Carry environment cwd values as `LegacyAppPathString` at the app-server
boundary and `PathUri` internally. Existing tool-call schemas and
relative-path behavior stay host-native; remaining local-only consumers
convert explicitly and leave follow-up TODOs.
The Wine integration test verifies app-server can start a thread and
complete an ordinary turn with a Windows environment cwd from Linux.
## Validation
- `bazel test //codex-rs/core/tests/remote_env_windows:smoke-test
--test_output=errors`
- focused app-server environment-selection and protocol schema tests
- scoped Clippy for `codex-core` and `codex-app-server-protocol`
## Why
Multi-agent v2 messages need a consistent, model-visible envelope that
identifies what kind of interaction occurred, who sent it, and which
agent it targets. Previously, encrypted deliveries exposed only
`encrypted_content`, while child completion used the legacy
`<subagent_notification>` shape. That meant the client could not
consistently present `NEW_TASK`, `MESSAGE`, and `FINAL_ANSWER` using the
same format.
This change adds the routing envelope as plaintext while keeping task
and message payloads encrypted. No new Responses API field is required:
an encrypted delivery is represented as an `input_text` header
immediately followed by its existing `encrypted_content` item.
Every envelope now follows this shape:
```text
Message Type: <NEW_TASK | MESSAGE | FINAL_ANSWER>
Task name: <recipient agent path>
Sender: <author agent path>
Payload:
<message payload>
```
## Message types
### `NEW_TASK`
`NEW_TASK` is used when the recipient should begin a new turn, including
an initial `spawn_agent` task and a later `followup_task`.
For a root agent spawning `/root/worker`, the request contains a
plaintext envelope followed by the encrypted task:
```json
{
"type": "agent_message",
"author": "/root",
"recipient": "/root/worker",
"content": [
{
"type": "input_text",
"text": "Message Type: NEW_TASK\nTask name: /root/worker\nSender: /root\nPayload:\n"
},
{
"type": "encrypted_content",
"encrypted_content": "<encrypted task payload>"
}
]
}
```
Conceptually, the model receives:
```text
Message Type: NEW_TASK
Task name: /root/worker
Sender: /root
Payload:
Review the authentication changes and report any regressions.
```
### `MESSAGE`
`MESSAGE` is used for a queued `send_message` delivery. It communicates
with an existing agent without starting a new turn.
For `/root/worker` reporting progress to the root agent, the request
contains:
```json
{
"type": "agent_message",
"author": "/root/worker",
"recipient": "/root",
"content": [
{
"type": "input_text",
"text": "Message Type: MESSAGE\nTask name: /root\nSender: /root/worker\nPayload:\n"
},
{
"type": "encrypted_content",
"encrypted_content": "<encrypted message payload>"
}
]
}
```
Conceptually, the model receives:
```text
Message Type: MESSAGE
Task name: /root
Sender: /root/worker
Payload:
The protocol tests pass; I am checking the resume path now.
```
### `FINAL_ANSWER`
`FINAL_ANSWER` is emitted when a child agent reaches a terminal state
and reports its result to its parent. Completion payloads are already
available locally, so the complete envelope is represented as plaintext
rather than as a plaintext header plus encrypted content.
For `/root/worker` completing work for the root agent, the request
contains:
```json
{
"type": "agent_message",
"author": "/root/worker",
"recipient": "/root",
"content": [
{
"type": "input_text",
"text": "Message Type: FINAL_ANSWER\nTask name: /root\nSender: /root/worker\nPayload:\nNo regressions found."
}
]
}
```
The model-visible form is:
```text
Message Type: FINAL_ANSWER
Task name: /root
Sender: /root/worker
Payload:
No regressions found.
```
Errored, shut down, and missing agents also use `FINAL_ANSWER`, with a
terminal-status description in the payload.
## What changed
- Render `NEW_TASK` or `MESSAGE` in
`InterAgentCommunication::to_model_input_item`, based on whether the
encrypted delivery starts a turn.
- Replace the multi-agent v2 `<subagent_notification>` completion
payload with a model-visible `FINAL_ANSWER` envelope.
- Document `Task name`, `Sender`, and `Payload` consistently in the
multi-agent developer instructions.
- Prevent local-only history projections from treating an encrypted
message's plaintext header as the complete assistant message.
- Preserve rollout-trace interaction edges when an agent message
contains both plaintext and encrypted content.
Legacy multi-agent behavior remains unchanged.
## Verification
- `just test -p codex-protocol`
- `just test -p codex-rollout-trace`
- `just test -p codex-web-search-extension`
- `just test -p codex-core
encrypted_multi_agent_v2_spawn_sends_agent_message_to_child`
- `just test -p codex-core
plaintext_multi_agent_v2_completion_sends_agent_message`
- `just test -p codex-core
multi_agent_v2_followup_task_completion_notifies_parent_on_every_turn`
- `just test -p codex-core
multi_agent_v2_completion_queues_message_for_direct_parent`
Summary
- Add the two missing `metadata: None` initializers after #28355 made
response-item metadata required.
- Restore test compilation for `codex-core` and `codex-api` on main.
Validation
- `git diff --check`
- `just fmt` (Rust formatting passed; unrelated Python formatter steps
could not use the sandboxed shared `uv` cache)
- Focused crate tests are running after PR creation.
## Description
This PR adds an optional `metadata` field to `ResponseItem` for
Responses API calls. Only mechanical plumbing, no actual values
populated and sent yet. Turns out just adding a new field to
`ResponseItem` has quite a large blast radius already.
This change is backwards compatible because `metadata` is optional and
omitted when absent, so existing response items and rollout history
without it still deserialize and requests that do not set it keep the
same wire shape. For provider compatibility, we strip out `metadata`
before non-OpenAI Responses requests so Azure and AWS Bedrock never see
this field.
My followup PR here will actually make use of it to start storing and
passing along `turn_id`: https://github.com/openai/codex/pull/28360
## What changed
- Added `ResponseItemMetadata` with optional `turn_id`, plus optional
`metadata` on Responses API item variants and inter-agent communication.
- Preserved item metadata through response-item rewrites such as
truncation, missing tool-output synthesis, compaction history
rebuilding, visible-history conversion, rollout/resume, and generated
app-server schemas/types.
- Strip item metadata from non-OpenAI Responses requests while
preserving it for OpenAI-shaped requests.
- Updated the mechanical fixture/test construction churn required by the
new optional field.
## Why
Guardian reviews embed the parent session transcript as untrusted
evidence. Skill or plugin mentions in that transcript must not be
interpreted as requests to inject more instructions into the Guardian
request, and memory context adds unrelated model-visible context to an
approval decision.
Keeping those sources out of the nested review session makes the request
smaller and preserves the trust boundary around the transcript being
assessed.
## What changed
- Skip skill and plugin discovery when building turns for Guardian
reviewer sessions.
- Disable memory context and dedicated memory tools in the derived
Guardian configuration.
- Extend the Guardian request-layout coverage to verify that a `$skill`
mention remains visible only as transcript evidence while neither the
skill body nor memory context is injected.
- Expand the Guardian configuration test to cover the disabled memory
settings.
## Testing
- Updated the Guardian review request snapshot and assertions for skill
and memory isolation.
- Extended the Guardian session configuration test to cover memories.
## Why
We need request-level evidence for Guardian cases where
`codex-auto-review` is missing from the client-side model catalog and
the review falls back to the parent model.
## What changed
- Add `guardian_catalog_contains_auto_review` to Guardian Responses API
client metadata.
- Add `guardian_model_provider_id` to Guardian Responses API client
metadata.
- Keep review-session metadata optional so callers without metadata
preserve the existing `None` path.
- Add tests for override, normal preferred-model, and
missing-auto-review-catalog behavior.
## Validation
- `just test -p codex-core
guardian_review_records_missing_auto_review_model_in_request_metadata`
- `just test -p codex-core
guardian_review_uses_model_catalog_override_when_preferred_review_model_exists`
- `just test -p codex-core
guardian_review_uses_preferred_review_model_without_model_catalog_override`
- `git diff --check origin/main`
## Why
Multi-agent v2 `send_message` deliveries already reach the receiving
model as typed `agent_message` items with encrypted content.
Child-completion notifications are generated by Codex itself, so their
content is plaintext and previously fell back to a serialized JSON
envelope inside an assistant message.
With plaintext `input_text` supported for `agent_message`, both delivery
paths can use the same model-visible type while preserving explicit
author and recipient metadata.
## What changed
- add plaintext `input_text` support to `AgentMessageInputContent` and
regenerate the affected app-server schemas
- preserve `InterAgentCommunication` as structured mailbox input instead
of converting it to assistant text
- record delivered communications as typed `agent_message` history items
- persist a dedicated rollout item so local delivery metadata such as
`trigger_turn` remains available without leaking into the Responses
request
- reconstruct typed agent messages on resume and preserve fork-turn
truncation behavior
- remove request-time assistant-content parsing
- preserve plaintext and encrypted inter-agent deliveries in stage-one
memory inputs
- normalize and link plaintext and encrypted agent messages in rollout
traces without treating inbound messages as child results
- cover the real MultiAgent V2 child-completion path end to end with
deterministic mailbox synchronization
## Verification
- `just test -p codex-core
plaintext_multi_agent_v2_completion_sends_agent_message`
- `just test -p codex-core input_queue_drains_mailbox_in_delivery_order
record_initial_history_reconstructs_typed_inter_agent_message
fork_turn_positions_use_inter_agent_delivery_metadata`
- `just test -p codex-memories-write
serializes_inter_agent_communications_for_memory`
- `just test -p codex-rollout-trace
agent_messages_preserve_routing_and_content
sub_agent_started_activity_creates_spawn_edge`
- `just test -p codex-rollout-trace
agent_result_edge_falls_back_to_child_thread_without_result_message`
- `just test -p codex-protocol -p codex-rollout -p
codex-app-server-protocol`
## Why
Sandbox cwd values can cross app-server and exec-server host boundaries.
They should retain URI semantics until the receiving host validates them
instead of being interpreted early as native paths.
## What
- Carry `PathUri` through filesystem sandbox contexts, sandbox commands,
and transform inputs.
- Convert command and policy cwd once in `SandboxManager::transform`,
then keep launch requests native.
- Preserve sandbox cwd over remote filesystem transport and reject
non-native URIs without fallback.
- Cache paired native/URI turn-environment cwd values during migration,
with immutable access to keep them synchronized.
- Extend existing protocol, forwarding, transform, and core runtime
tests.
## Why
We want to remove implicit use of `$CODEX_HOME` from `codex-core` and
make embedders responsible for supplying user-level instructions. This
also ensures user instructions load when no primary environment is
selected.
## What changed
Stacked on #27415, which makes `codex exec` surface thread-scoped
runtime warnings.
- Added `UserInstructionsProvider` to `codex-extension-api`, with
absolute source attribution and recoverable loading warnings.
- Added `codex-home` with the filesystem-backed provider for
`AGENTS.override.md` and `AGENTS.md`, preserving precedence, fallback,
trimming, lossy UTF-8 handling, and the existing uncapped global
instruction size.
- Removed global instruction loading from `Config` and require
`ThreadManager` callers to inject a provider.
- Load provider instructions once for each fresh root runtime, including
runtimes without a primary environment. Running sessions retain their
snapshot, while child agents inherit the parent snapshot without
invoking the provider.
- Keep provider instructions separate while loading project `AGENTS.md`,
then assemble the model-visible instructions with the existing ordering,
source attribution, warning, and turn-context behavior.
- Wired the Codex home provider through the CLI, app server, MCP server,
core facade, and thread-manager sample.
## Validation
- `just test -p codex-home -p codex-extension-api`
- `just test -p codex-core agents_md`
- `just test -p codex-core guardian`
- `just test -p codex-app-server
thread_start_without_selected_environment_includes_only_global_instruction_source`
- `just test -p codex-exec warning`
- `just bazel-lock-check`
## Background
Codex can use **Auto Review** for permission requests. Instead of asking
the user immediately, Codex starts a separate locked-down reviewer
session called **Guardian**, which returns a structured `allow` or
`deny` assessment.
The Guardian reviewer is itself a Codex session, so its model request
can fail for transient infrastructure reasons such as model overload,
HTTP connection failure, or response-stream disconnect. Today, any such
failure immediately ends the Auto Review attempt and blocks the action.
This PR adds bounded retries for failures that the existing protocol
explicitly identifies as transient.
Linear context:
[CA-539](https://linear.app/openai/issue/CA-539/retry-auto-review-infrastructure-failures-and-fall-back-to-manual)
## What changes
A Guardian review can now make at most **three total attempts**:
1. Run the review normally.
2. Retry after a jittered delay of roughly 180–220 ms if the first
attempt fails with an eligible error.
3. Retry after a jittered delay of roughly 360–440 ms if the second
attempt also fails with an eligible error.
All attempts share the original review deadline. Jitter spreads retries
from concurrent clients to reduce synchronized load during broader
outages. The retries do not reset the user's maximum wait time, and the
backoff waits terminate early if the review is cancelled or the deadline
expires.
Before retrying, the existing Guardian session lifecycle decides whether
the session remains usable. Healthy trunks are reused, broken trunks are
removed by the existing cleanup path, and ephemeral sessions continue to
clean themselves up.
The review still emits one logical lifecycle to clients. Recoverable
intermediate failures do not produce warnings or terminal events.
## Retry policy
### Retried up to twice
- model/server overload
- HTTP connection failure
- response-stream connection failure
- response-stream disconnect
- internal server error
- a final reviewer message that cannot be parsed as the required
Guardian assessment
### Not retried
- bad or invalid requests
- authentication failures
- usage limits
- cyber-policy failures
- errors without a structured category
- a request that already exhausted the lower-level Responses retry
budget
- a completed Guardian turn with no assessment payload
- prompt-construction failures
- Guardian review timeout
- cancellation or abort
- a valid `deny` assessment
The session-error classification uses `ErrorEvent.codex_error_info`; it
does not inspect error-message strings.
## Implementation notes
- `wait_for_guardian_review` preserves the complete `ErrorEvent`,
including structured `codex_error_info`.
- Guardian session failures preserve the original message and optional
structured `CodexErrorInfo`.
- The retry policy classifies the explicitly transient `CodexErrorInfo`
variants; unknown, absent, and deterministic categories are not retried.
- The Guardian session manager receives the caller's deadline rather
than creating a new timeout per attempt.
- Analytics record the final `attempt_count`.
- Retry orchestration does not add a separate session-cleanup protocol;
it relies on the existing trunk and ephemeral lifecycle decisions.
## Automated testing
Focused Guardian coverage verifies:
- every supported transient `CodexErrorInfo` is classified as retryable,
while absent and non-transient categories are not;
- structured transient session failure -> retry -> approval with the
healthy trunk reused;
- two invalid Guardian responses -> third attempt -> approval, with
exactly three requests;
- three invalid responses -> existing fail-closed result, with exactly
three requests and one terminal lifecycle;
- valid denial, missing payload, invalid request, timeout, cancellation,
and prompt/session construction failures are not retried;
- retry eligibility ends after the third attempt;
- retry delays use the shared exponential backoff helper and remain
within the expected jitter bounds;
- cancellation and deadline expiry interrupt the backoff wait;
- healthy trunks are reused across retryable failures;
- broken event streams remove the trunk through the existing lifecycle
cleanup;
- an ephemeral retry does not disturb a concurrent trunk review.
Validation performed:
- `just test -p codex-core guardian_review_
guardian_ephemeral_retry_preserves_parallel_trunk_and_fork_history
run_review_removes_trunk_when_event_stream_is_broken` — **42 passed**;
- `just test -p codex-analytics` — **71 passed**;
- scoped Clippy fixes for `codex-core` and `codex-analytics` passed.
A prior full `codex-core` run had unrelated environment-sensitive
failures outside Guardian coverage.
## Manual QA
The focused integration tests use the local mock Responses server to
inspect exact request counts and emitted lifecycle events. They confirm
that retries are internal, a successful later attempt supplies the final
decision, non-retryable failures issue only one request, and exhausted
retries emit only one terminal result.
## Why
Thread cwd and environment selections are a single logical setting in
core: updating one without the other can silently desynchronize the
next-turn execution context. This change makes that relationship
explicit in the internal thread settings flow while preserving the
existing app-server public API shape.
## What changed
- Moved the cwd/environment pair through internal
`ThreadSettingsOverrides.environment_settings` instead of a top-level
internal `cwd` field.
- Kept `thread/settings/update` public params unchanged, with app-server
translating top-level `cwd` into the paired internal settings shape.
- Moved `Op::UserInput` environment overrides into thread settings so
user turns and settings updates use the same core path.
- Updated core, app-server, MCP, memories, sample, and test callsites to
construct the paired settings shape.
## Verification
- `git diff --check`
- Local test run starting after PR creation.
## Why
Thread settings cwd overrides are expected to be resolved before they
enter core. Keeping this boundary as a plain `PathBuf` made it easy for
core/session code to keep fallback normalization and relative-path
resolution logic in places that should only receive an already-resolved
cwd.
This is intentionally the absolute-cwd-only slice: it does not change
environment selection stickiness or cwd-to-default-environment fallback
behavior.
## What changed
- Changes `ThreadSettingsOverrides.cwd`,
`CodexThreadSettingsOverrides.cwd`, and `SessionSettingsUpdate.cwd` to
use `AbsolutePathBuf`.
- Removes core-side cwd normalization/resolution from session settings
updates.
- Updates affected core/app-server test helpers and callsites to pass
existing absolute cwd values or use `abs()` helpers.
## Validation
Opening as draft so CI can start while local validation continues.
rhan-oai