sayan-oai dd43a9967f Use step environments for Guardian approval reviews (#37618) ...

## Why

Deferred environments can become ready after a turn starts. Guardian approval
reviews must use the environment selected for the current step so their working
directory and permission context are not taken from the stale turn snapshot.

## What changed

- Carry `StepContext` through tool execution and approval handling.
- Add a Guardian review context that pairs the parent turn with the current
  step's environment snapshot.
- Use that snapshot for Guardian prompts, reviewer sessions, and delegated
  reviewer threads while retaining turn-level fallbacks for callers without a
  step context.

## Testing

Added an integration test that makes a deferred remote environment ready before
an escalated command and verifies Guardian receives that environment's working
directory and denied-read policy instead of the initial local environment's.

GitOrigin-RevId: 779096bc7d0e0c7453a009c02141029714b9b407

2026-08-08 20:07:27 +00:00

..

snapshots

Include sandbox mode in response metadata (#37507)

2026-08-07 22:41:20 +00:00

abort_tasks.rs

Pair thread environment settings (#26687)

2026-06-08 13:55:15 -07:00

additional_context.rs

Pair thread environment settings (#26687)

2026-06-08 13:55:15 -07:00

agent_execution.rs

Support leaf models in multi-agent v2 (#36892)

2026-08-04 12:36:14 +00:00

agent_websocket.rs

Remove the legacy code-mode tool metadata inventory (#37500)

2026-08-07 20:26:08 +00:00

agents_md.rs

Cap project instructions across environments (#37424)

2026-08-07 09:50:43 +00:00

apply_patch_cli.rs

Render turn diffs for foreign environment paths (#34654)

2026-07-22 02:19:45 +00:00

approvals.rs

Avoid reinjecting permissions after command approvals (#36800)

2026-08-03 21:37:17 +00:00

audio_truncation.rs

Consolidate thread startup around StartThreadOptions (#34814)

2026-07-22 19:31:54 +00:00

auto_review.rs

Gate Apps usage instructions by model capability (#37145)

2026-08-05 20:03:13 +00:00

catalog_permission_messages.rs

Use model catalog collaboration mode messages (#36351)

2026-07-31 15:40:52 +00:00

cli_stream.rs

Update models.json (#31684)

2026-07-09 03:40:08 +00:00

client_websockets.rs

Preserve base instruction provenance across sessions (#37446)

2026-08-07 14:08:13 +00:00

client.rs

Send model routing hints to the Codex backend (#37345)

2026-08-07 00:58:54 +00:00

code_mode_elicitation.rs

code-mode: make all approvals trigger elicitation pause (#31650)

2026-07-08 15:27:04 -07:00

code_mode.rs

Remove the legacy code-mode tool metadata inventory (#37500)

2026-08-07 20:26:08 +00:00

codex_delegate.rs

Remove obsolete ignored tests (#34558)

2026-07-21 14:57:39 +00:00

collaboration_instructions.rs

Enforce single-writer ownership for all thread histories (#36389)

2026-07-31 20:49:08 +00:00

compact_remote_parity.rs

Always assign response item IDs (#34645)

2026-07-22 01:23:44 +00:00

compact_remote.rs

Include sandbox mode in response metadata (#37507)

2026-08-07 22:41:20 +00:00

compact_resume_fork.rs

Negotiate MCP extensions per app-server session (#36910)

2026-08-04 15:23:43 +00:00

compact.rs

Enforce single-writer ownership for all thread histories (#36389)

2026-07-31 20:49:08 +00:00

current_time_reminder.rs

Refresh environment dates from the configured clock (#36187)

2026-07-30 16:44:18 +00:00

cyber_exec_policy.rs

Ignore reusable command approvals for cyber models (#37516)

2026-08-08 00:59:39 +00:00

deprecation_notice.rs

Add indexed web search mode (#28489)

2026-06-19 05:35:57 -07:00

exec_policy.rs

Propagate approval rejection reasons (#34400)

2026-07-20 19:23:21 +00:00

exec.rs

Add network environment ID plumbing (#28766)

2026-06-18 14:09:38 +02:00

extension_sandbox.rs

Skip missing paths in filesystem sandbox entries (#34598)

2026-07-21 19:17:18 +00:00

external_auth.rs

Remove codex-core's direct reqwest dependency (#36076)

2026-07-30 04:03:33 +00:00

fork_thread.rs

Negotiate MCP extensions per app-server session (#36910)

2026-08-04 15:23:43 +00:00

git_enrichment.rs

Stabilize the startup Git enrichment integration test (#36769)

2026-08-03 15:59:00 +00:00

guardian_review.rs

Interrupt cyber model turns after one Guardian denial (#37190)

2026-08-06 00:37:46 +00:00

hooks_mcp.rs

Add per-surface MCP tool exposure controls (#36781)

2026-08-03 18:51:46 +00:00

hooks.rs

Support asynchronous command hooks (#37533)

2026-08-08 04:24:12 +00:00

image_rollout.rs

Report prompt image resizing to the model (#37134)

2026-08-05 18:19:22 +00:00

injected_models_cache.rs

Allow injecting model catalog caches (#36992)

2026-08-04 23:04:11 +00:00

items.rs

Include item start times in completion events (#35363)

2026-07-25 16:06:54 +00:00

json_result.rs

Pair thread environment settings (#26687)

2026-06-08 13:55:15 -07:00

live_cli.rs

[codex] Use expect in integration tests (#28441)

2026-06-15 21:53:47 -07:00

mcp_auth_elicitation.rs

Add keyed shell environment policy filters (#34590)

2026-07-21 18:23:26 +00:00

mcp_auth_refresh.rs

Unify plugin skill loading with the host skill service (#37444)

2026-08-07 13:54:24 +00:00

mcp_refresh_cleanup.rs

Negotiate MCP extensions per app-server session (#36910)

2026-08-04 15:23:43 +00:00

mcp_startup_refresh_http_proxy.rs

Add per-surface MCP tool exposure controls (#36781)

2026-08-03 18:51:46 +00:00

mcp_tool_cache.rs

Reuse MCP handlers across sampling steps (#37273)

2026-08-06 13:37:07 +00:00

mcp_tool_exposure.rs

Expose session sources to MCP contributors (#37167)

2026-08-05 21:45:33 +00:00

mcp_turn_metadata.rs

Enforce automatic review for managed models (#37511)

2026-08-07 23:43:14 +00:00

mod.rs

Ignore reusable command approvals for cyber models (#37516)

2026-08-08 00:59:39 +00:00

model_overrides.rs

[5 of 7] Replace OverrideTurnContext with ThreadSettings (#22508)

2026-05-18 21:03:51 -07:00

model_runtime_selectors.rs

Run code mode exclusively through the standalone host (#36217)

2026-07-30 20:24:29 +00:00

model_switching.rs

Preserve base instruction provenance across sessions (#37446)

2026-08-07 14:08:13 +00:00

model_visible_layout.rs

Preserve foreign cwd URIs for turn-input contributors (#37342)

2026-08-07 00:42:06 +00:00

models_cache_ttl.rs

Gate Apps usage instructions by model capability (#37145)

2026-08-05 20:03:13 +00:00

models_etag_responses.rs

Pair thread environment settings (#26687)

2026-06-08 13:55:15 -07:00

multi_agent_mode.rs

Enforce single-writer ownership for all thread histories (#36389)

2026-07-31 20:49:08 +00:00

multi_agent_resume.rs

Read turn permissions from the current configuration (#36930)

2026-08-04 17:13:13 +00:00

multi_exec_server_sandbox.rs

Test workspace write isolation across exec servers (#33861)

2026-07-17 16:07:24 +00:00

network_approval.rs

Test remote Guardian network approval scoping (#36177)

2026-07-30 15:50:04 +00:00

openai_file_mcp.rs

Resolve MCP file uploads with environment-native paths (#36002)

2026-07-29 18:06:14 +00:00

otel.rs

Propagate approval rejection reasons (#34400)

2026-07-20 19:23:21 +00:00

override_updates.rs

Pair thread environment settings (#26687)

2026-06-08 13:55:15 -07:00

pending_input.rs

Accept user input when starting idle turns (#36947)

2026-08-04 18:52:41 +00:00

permissions_messages.rs

Consolidate model instructions in ModelMessages (#36787)

2026-08-03 19:30:40 +00:00

personality.rs

Gate Apps usage instructions by model capability (#37145)

2026-08-05 20:03:13 +00:00

plugins.rs

Unify plugin skill loading with the host skill service (#37444)

2026-08-07 13:54:24 +00:00

prompt_cache_key.rs

Use session IDs for prompt cache keys (#33035)

2026-07-14 10:37:23 +00:00

prompt_caching.rs

Consolidate model instructions in ModelMessages (#36787)

2026-08-03 19:30:40 +00:00

prompt_debug_tests.rs

Enable git attribution across Codex entry points (#34819)

2026-07-22 20:23:35 +00:00

quota_exceeded.rs

Pair thread environment settings (#26687)

2026-06-08 13:55:15 -07:00

realtime_conversation.rs

Negotiate MCP extensions per app-server session (#36910)

2026-08-04 15:23:43 +00:00

realtime_initial_items.rs

Add a realtime delegation acknowledgement control (#36413)

2026-08-01 01:17:26 +00:00

remote_env.rs

Use step environments for Guardian approval reviews (#37618)

2026-08-08 20:07:27 +00:00

remote_models.rs

Gate Apps usage instructions by model capability (#37145)

2026-08-05 20:03:13 +00:00

request_compression.rs

Pair thread environment settings (#26687)

2026-06-08 13:55:15 -07:00

request_permissions_tool.rs

Isolate apply-patch permission test variants (#36138)

2026-07-30 11:41:47 +00:00

request_permissions.rs

Propagate approval rejection reasons (#34400)

2026-07-20 19:23:21 +00:00

request_plugin_install.rs

Respect plugin skill availability in tool suggestions (#37447)

2026-08-07 14:17:14 +00:00

request_user_input.rs

Remove legacy collaboration mode variants (#36990)

2026-08-04 22:58:02 +00:00

responses_api_proxy_headers.rs

Pair thread environment settings (#26687)

2026-06-08 13:55:15 -07:00

responses_lite.rs

Remove the legacy code-mode tool metadata inventory (#37500)

2026-08-07 20:26:08 +00:00

responses_system_proxy.rs

Migrate login HTTP construction to HttpClient (#34643)

2026-07-22 00:50:28 +00:00

resume_warning.rs

Negotiate MCP extensions per app-server session (#36910)

2026-08-04 15:23:43 +00:00

resume.rs

Enforce single-writer ownership for all thread histories (#36389)

2026-07-31 20:49:08 +00:00

review.rs

Use current session settings for review threads (#36941)

2026-08-04 18:09:58 +00:00

rmcp_client.rs

Recover MCP servers after OAuth reauthentication (#37337)

2026-08-07 00:17:46 +00:00

rollout_budget.rs

Use provider-reported rollout budget units (#36715)

2026-08-03 09:46:38 +00:00

rollout_list_find.rs

Honor the configured SQLite home across state consumers (#34994)

2026-07-23 19:19:35 +00:00

safety_buffering.rs

[codex] Read retry model from buffering events (#31262)

2026-07-06 17:04:26 +00:00

safety_check_downgrade.rs

Update models.json (#21818)

2026-07-08 16:24:12 -07:00

search_tool.rs

Canonicalize default tools under the functions namespace (#37022)

2026-08-05 04:07:59 +00:00

shell_command.rs

Require explicit sandbox permissions with shell justifications (#36350)

2026-07-31 15:34:03 +00:00

shell_serialization.rs

test: branch on target OS instead of runner flavor (#29712)

2026-06-23 14:27:13 -07:00

shell_snapshot.rs

Wait for patch events before reading shell snapshots (#36757)

2026-08-03 14:14:41 +00:00

skill_approval.rs

Pair thread environment settings (#26687)

2026-06-08 13:55:15 -07:00

skills_extension.rs

Move host skill prompt injection into the skills extension (#37503)

2026-08-07 21:04:16 +00:00

skills.rs

Preserve discovery paths for symlinked skills (#37144)

2026-08-05 19:51:51 +00:00

spawn_agent_description.rs

Track multi-agent usage hints in world state (#37189)

2026-08-06 00:22:06 +00:00

sqlite_state.rs

Add per-surface MCP tool exposure controls (#36781)

2026-08-03 18:51:46 +00:00

stream_error_allows_next_turn.rs

Allow custom providers to opt into standalone web search (#34846)

2026-07-23 00:02:15 +00:00

stream_no_completed.rs

Keep response streams alive through connection failures (#37485)

2026-08-07 18:33:46 +00:00

subagent_notifications.rs

Track context windows per agent (#37347)

2026-08-07 01:05:33 +00:00

token_budget.rs

Preserve base instruction provenance across sessions (#37446)

2026-08-07 14:08:13 +00:00

tool_harness.rs

[codex] Use expect in integration tests (#28441)

2026-06-15 21:53:47 -07:00

tool_parallelism.rs

[codex] Use expect in integration tests (#28441)

2026-06-15 21:53:47 -07:00

tools.rs

Reject conflicting namespace descriptions in strict tool mode (#37053)

2026-08-05 08:01:47 +00:00

truncation.rs

Add per-surface MCP tool exposure controls (#36781)

2026-08-03 18:51:46 +00:00

turn_state.rs

Include sandbox mode in response metadata (#37507)

2026-08-07 22:41:20 +00:00

unified_exec_process_events.rs

Delegate remote process sandboxing to the executor (#37480)

2026-08-07 18:05:17 +00:00

unified_exec_zsh_fork_approvals.rs

Skip missing paths in filesystem sandbox entries (#34598)

2026-07-21 19:17:18 +00:00

unified_exec.rs

Honor per-environment login shell policy (#36811)

2026-08-03 23:33:05 +00:00

unstable_features_warning.rs

Negotiate MCP extensions per app-server session (#36910)

2026-08-04 15:23:43 +00:00

user_message_admission.rs

Add durable user-message queue dispatch (#37204)

2026-08-06 03:11:42 +00:00

user_notification.rs

Pair thread environment settings (#26687)

2026-06-08 13:55:15 -07:00

user_shell_cmd.rs

[codex] Warn clearly when code mode output is truncated (#28467)

2026-06-16 10:37:06 -07:00

view_image.rs

Add a unified image budget (#37206)

2026-08-06 03:15:08 +00:00

web_search.rs

Enable cached web search for Amazon Bedrock (#36938)

2026-08-04 17:55:28 +00:00

websocket_fallback.rs

Pair thread environment settings (#26687)

2026-06-08 13:55:15 -07:00

window_headers.rs

[codex] Use expect in integration tests (#28441)

2026-06-15 21:53:47 -07:00

windows_sandbox.rs

Skip missing paths in filesystem sandbox entries (#34598)

2026-07-21 19:17:18 +00:00

workspace_roots.rs

Use turn environment permissions for tool execution (#37038)

2026-08-05 06:38:57 +00:00