jif 5d3fe48b08 Improve Guardian retries and review failure reporting (#44482) ...

## Why

Transient rate limits can end automatic approval reviews prematurely, and review failures currently report high risk even when no assessment completed.

## What changed

- Retry rate limits and recoverable exhausted-stream errors, while excluding non-transient HTTP failures.
- Preserve server retry delays after stream retries are exhausted and honor them within the review deadline. Scope retry advice to the current turn so reused sessions cannot apply stale delays.
- Keep failed reviews denied, but leave risk and authorization unset and explain that the review could not complete without declaring the action unsafe.

## Testing

Add an integration test covering rate-limit recovery through approval and tool execution, asserting that the action executes exactly once after approval. Update failure assertions to check absent assessment fields and the review-failure explanation.

GitOrigin-RevId: 1163cfde35c6b8eb23b6f24f1f24461b86ded838

2026-09-10 09:50:59 +00:00

..

agent

Persist disabled plugin IDs in thread settings (#44332)

2026-09-09 23:23:59 +00:00

apps

…

config

Give hosted Codex Apps an independent MCP protocol opt-in (#44318)

2026-09-09 22:22:08 +00:00

context

Surface environment startup failure reasons to the model (#44277)

2026-09-09 18:50:32 +00:00

context_manager

Persist disabled plugin IDs in thread settings (#44332)

2026-09-09 23:23:59 +00:00

exec_policy

Apply remote platform semantics to exec safety checks (#42113)

2026-09-01 18:24:56 +00:00

guardian

Improve Guardian retries and review failure reporting (#44482)

2026-09-10 09:50:59 +00:00

mcp_tool_call

Use captured step settings for tool planning and execution (#44242)

2026-09-09 17:45:37 +00:00

plugins

Include originator in plugin measurement analytics (#42445)

2026-09-03 04:38:26 +00:00

realtime_conversation

Support attaching to existing realtime calls (#39876)

2026-08-21 06:41:28 +00:00

realtime_history

Record realtime conversation history in Core (#41924)

2026-08-31 22:38:23 +00:00

sandboxing

Derive sandbox contexts from turn environments (#40551)

2026-08-25 05:16:45 +00:00

session

Harden Code Mode tool-call completeness tracking (#44472)

2026-09-10 08:57:35 +00:00

state

Preserve prewarmed reasoning effort across replay and early rollback (#44285)

2026-09-09 19:38:24 +00:00

tasks

Always use streamed remote compaction for supported providers (#44255)

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

tools

Harden Code Mode tool-call completeness tracking (#44472)

2026-09-10 08:57:35 +00:00

unified_exec

Use captured step settings for tool planning and execution (#44242)

2026-09-09 17:45:37 +00:00

utils

Count serialized JSON bytes without allocating strings (#40604)

2026-08-25 13:10:24 +00:00

agent_communication.rs

…

agents_md_manager.rs

Rename host instruction payload to Instructions (#40709)

2026-08-25 21:20:51 +00:00

agents_md_tests.rs

Honor environment-resolved workspace roots (#40912)

2026-08-26 16:28:35 +00:00

agents_md.rs

Add tracing for project instructions and filesystem sandbox operations (#43913)

2026-09-08 21:11:43 +00:00

apply_patch_tests.rs

Authorize apply_patch in the executor path context (#42391)

2026-09-02 23:58:24 +00:00

apply_patch.rs

Authorize apply_patch in the executor path context (#42391)

2026-09-02 23:58:24 +00:00

attestation.rs

…

client_common_tests.rs

Normalize image detail for the receiving model (#44249)

2026-09-09 17:56:06 +00:00

client_common.rs

Normalize image detail for the receiving model (#44249)

2026-09-09 17:56:06 +00:00

client_tests.rs

Add bounded tool-result metadata support to executed tool calls (#44336)

2026-09-09 23:46:17 +00:00

client.rs

Add bounded tool-result metadata support to executed tool calls (#44336)

2026-09-09 23:46:17 +00:00

codex_delegate_tests.rs

Distinguish Guardian review threads from subagents (#40221)

2026-08-23 10:06:57 +00:00

codex_delegate.rs

Persist disabled plugin IDs in thread settings (#44332)

2026-09-09 23:23:59 +00:00

codex_thread.rs

Persist disabled plugin IDs in thread settings (#44332)

2026-09-09 23:23:59 +00:00

command_canonicalization_tests.rs

…

command_canonicalization.rs

…

compact_model_fallback.rs

Remove the unused legacy remote compaction implementation (#44273)

2026-09-09 18:40:53 +00:00

compact_remote_history_tests.rs

Remove the unused legacy remote compaction implementation (#44273)

2026-09-09 18:40:53 +00:00

compact_remote_history.rs

Remove the unused legacy remote compaction implementation (#44273)

2026-09-09 18:40:53 +00:00

compact_remote_v2_attempt.rs

Remove the unused legacy remote compaction implementation (#44273)

2026-09-09 18:40:53 +00:00

compact_remote_v2_image_budget_tests.rs

Support per-tool MCP output limits (#41421)

2026-08-28 21:40:23 +00:00

compact_remote_v2_images.rs

Budget retained images during remote compaction (#40280)

2026-08-23 21:29:11 +00:00

compact_remote_v2.rs

Remove the unused legacy remote compaction implementation (#44273)

2026-09-09 18:40:53 +00:00

compact_tests.rs

Remove the unused legacy remote compaction implementation (#44273)

2026-09-09 18:40:53 +00:00

compact_token_budget.rs

Snapshot resolved settings for each model step (#40651)

2026-08-25 17:41:22 +00:00

compact.rs

Remove the unused legacy remote compaction implementation (#44273)

2026-09-09 18:40:53 +00:00

connectors_tests.rs

Honor app link settings for MCP tool approvals (#42056)

2026-09-01 13:43:27 +00:00

connectors.rs

Honor app link settings for MCP tool approvals (#42056)

2026-09-01 13:43:27 +00:00

current_time.rs

…

cyber_access_program.rs

Add per-turn cyber access program selection (#40687)

2026-08-25 19:54:52 +00:00

elicitation_tests.rs

…

elicitation.rs

…

environment_selection.rs

Surface environment startup failure reasons to the model (#44277)

2026-09-09 18:50:32 +00:00

event_mapping_tests.rs

Identify agents by name in token budget context (#36815)

2026-08-04 00:51:53 +00:00

event_mapping.rs

Add structured asynchronous user input requests (#42178)

2026-09-02 02:09:35 +00:00

exec_env_tests.rs

Add a feature flag to preserve apply_patch line endings (#37758)

2026-08-10 01:44:30 +00:00

exec_env.rs

Expose the Codex version to commands and turn metadata (#42395)

2026-09-03 00:17:19 +00:00

exec_policy_tests.rs

Remove selected core test cases (#42066)

2026-09-01 14:43:27 +00:00

exec_policy_windows_tests.rs

Retire the untrusted approval policy (#39630)

2026-08-20 07:03:02 +00:00

exec_policy.rs

Apply remote platform semantics to exec safety checks (#42113)

2026-09-01 18:24:56 +00:00

exec_tests.rs

Harden credential handling in shell snapshots and replay (#44040)

2026-09-09 04:00:10 +00:00

exec.rs

Harden credential handling in shell snapshots and replay (#44040)

2026-09-09 04:00:10 +00:00

function_tool.rs

…

git_info_tests.rs

Sanitize credentials from Git remote metadata (#40713)

2026-08-25 21:34:05 +00:00

guardian_review.rs

Move synchronous Guardian orchestration into the reviewer extension (#44252)

2026-09-09 17:59:10 +00:00

hook_mcp_executor.rs

Preserve executor context for MCP stop hooks (#40012)

2026-08-21 21:17:42 +00:00

hook_runtime.rs

Distinguish forked sessions in session-start hooks (#44349)

2026-09-10 00:26:30 +00:00

image_preparation_tests.rs

Preserve content kinds during image preparation (#40281)

2026-08-23 22:23:40 +00:00

image_preparation.rs

Centralize prompt image detail modes (#42624)

2026-09-03 22:22:48 +00:00

installation_id.rs

…

lib.rs

Add telemetry for the Windows system config namespace (#44284)

2026-09-09 19:31:12 +00:00

mcp_openai_file.rs

Honor environment-resolved workspace roots (#40912)

2026-08-26 16:28:35 +00:00

mcp_skill_dependencies.rs

Preserve acceptance order in retained thread context (#42770)

2026-09-04 13:50:27 +00:00

mcp_tool_approval_templates.rs

Organize bundled Rust resources under asset directories (#41477)

2026-08-29 06:05:28 +00:00

mcp_tool_call_tests.rs

Scope session MCP approvals to app account links (#42133)

2026-09-01 20:46:17 +00:00

mcp_tool_call.rs

Use captured step settings for tool planning and execution (#44242)

2026-09-09 17:45:37 +00:00

mcp_tool_exposure_test.rs

Reuse MCP handlers across sampling steps (#37273)

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

mcp_tool_exposure.rs

Honor app link settings for MCP tool approvals (#42056)

2026-09-01 13:43:27 +00:00

mcp.rs

Honor managed plugin disablement for executor capabilities (#40640)

2026-08-25 16:47:13 +00:00

memory_usage.rs

Standardize shell execution on unified exec (#39757)

2026-08-20 17:46:05 +00:00

mention_syntax.rs

…

network_policy_decision_tests.rs

…

network_policy_decision.rs

…

original_image_detail.rs

…

otel_init.rs

Prepare the telemetry shutdown worker during initialization (#39050)

2026-08-17 19:06:53 +00:00

prompt_debug.rs

Use the originating model when recording conversation history (#44243)

2026-09-09 17:46:00 +00:00

realtime_context_tests.rs

Persist Daybreak preferences in thread metadata (#42854)

2026-09-04 21:17:25 +00:00

realtime_context.rs

Add experimental app-server project APIs (#38940)

2026-08-17 04:01:37 +00:00

realtime_conversation_tests.rs

Reconnect WebRTC Realtime sideband transports (#39257)

2026-08-18 18:41:40 +00:00

realtime_conversation.rs

Gate new turn submissions on host shutdown admission (#43943)

2026-09-08 23:45:10 +00:00

realtime_history_tests.rs

Add structured asynchronous user input requests (#42178)

2026-09-02 02:09:35 +00:00

realtime_history.rs

Record realtime conversation history in Core (#41924)

2026-08-31 22:38:23 +00:00

realtime_prompt.rs

…

responses_metadata.rs

Remove the unused legacy remote compaction implementation (#44273)

2026-09-09 18:40:53 +00:00

responses_retry_tests.rs

…

responses_retry.rs

Improve Guardian retries and review failure reporting (#44482)

2026-09-10 09:50:59 +00:00

rollout_budget.rs

Use provider-reported rollout budget units (#36715)

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

rollout.rs

…

safety_tests.rs

Authorize apply_patch in the executor path context (#42391)

2026-09-02 23:58:24 +00:00

safety.rs

Authorize apply_patch in the executor path context (#42391)

2026-09-02 23:58:24 +00:00

sandbox_tags_tests.rs

Report configured sandbox policy consistently (#41933)

2026-08-31 23:29:20 +00:00

sandbox_tags.rs

Report configured sandbox policy consistently (#41933)

2026-08-31 23:29:20 +00:00

session_prefix_tests.rs

…

session_prefix.rs

Classify additional generated context fragments (#40275)

2026-08-23 20:08:40 +00:00

session_rollout_init_error.rs

Preserve sandbox errors during session initialization (#40381)

2026-08-24 10:56:54 +00:00

session_startup_prewarm.rs

Move synchronous Guardian orchestration into the reviewer extension (#44252)

2026-09-09 17:59:10 +00:00

shell_snapshot_sandbox_tests.rs

Cache protected shell snapshots and harden capture cleanup (#43954)

2026-09-09 00:42:57 +00:00

shell_snapshot_sandbox.rs

Harden credential handling in shell snapshots and replay (#44040)

2026-09-09 04:00:10 +00:00

shell_snapshot_tests.rs

Support configured credential providers across shell snapshots (#44072)

2026-09-09 06:04:59 +00:00

shell_snapshot.rs

Support configured credential providers across shell snapshots (#44072)

2026-09-09 06:04:59 +00:00

shell_tests.rs

Resolve model-provided shells by type (#39607)

2026-08-20 06:03:25 +00:00

shell.rs

Resolve model-provided shells by type (#39607)

2026-08-20 06:03:25 +00:00

skills.rs

Trust invoked user skills in Guardian reviews (#41006)

2026-08-27 00:07:26 +00:00

spawn.rs

Support sandboxed file streaming in exec-server (#38356)

2026-08-13 11:11:26 +00:00

state_db_bridge.rs

…

stream_events_utils_tests.rs

Use the originating model when recording conversation history (#44243)

2026-09-09 17:46:00 +00:00

stream_events_utils.rs

Use the originating model when recording conversation history (#44243)

2026-09-09 17:46:00 +00:00

test_support.rs

Add bounded tool-result metadata support to executed tool calls (#44336)

2026-09-09 23:46:17 +00:00

thread_manager_tests.rs

Use captured step settings when building model context (#44200)

2026-09-09 16:48:03 +00:00

thread_manager.rs

Add opt-in model discovery for OpenAI API keys (#44392)

2026-09-10 03:18:28 +00:00

thread_rollout_truncation_tests.rs

Use captured step settings when building model context (#44200)

2026-09-09 16:48:03 +00:00

thread_rollout_truncation.rs

Preserve harness metadata across conversation history (#38058)

2026-08-11 19:30:12 +00:00

turn_diff_tracker_tests.rs

…

turn_diff_tracker.rs

…

turn_metadata_tests.rs

Set turn triggers for guardian and memory requests (#44298)

2026-09-09 20:32:57 +00:00

turn_metadata.rs

Set turn triggers for guardian and memory requests (#44298)

2026-09-09 20:32:57 +00:00

turn_timing_tests.rs

Replace Guardian tickets with parent response IDs (#43002)

2026-09-05 13:02:49 +00:00

turn_timing.rs

Propagate response tickets to Guardian reviews (#42758)

2026-09-04 12:56:26 +00:00

user_shell_command_tests.rs

…

user_shell_command.rs

Snapshot resolved settings for each model step (#40651)

2026-08-25 17:41:22 +00:00

util_tests.rs

…

util.rs

…

web_search.rs

…

windows_sandbox_read_grants_tests.rs

…

windows_sandbox_read_grants.rs

…

windows_sandbox_tests.rs

…

windows_sandbox.rs

Add experimental Windows sandbox service provisioning (#42353)

2026-09-02 20:30:45 +00:00

windows_system_config.rs

Add telemetry for the Windows system config namespace (#44284)

2026-09-09 19:31:12 +00:00

worktree_trust_tests.rs

Validate linked worktrees before inheriting project trust (#39616)

2026-08-20 06:29:32 +00:00