Commit Graph
4204 Commits
Author SHA1 Message Date
felixxia-oai 9949c9eafa Reuse Guardian reviews in async risk scoring (#40013)
## What changed

- Retain bounded evidence from completed synchronous Guardian allow and deny reviews and supply it to subsequent Guardian v2 async classifier samples as trusted developer context.
- Keep review evidence isolated from the conversation transcript, escape and truncate its fields, and ignore failed or incomplete reviews.
- Invalidate retained evidence after conversation history rewrites or new user messages, including authorization changes in a worker's root thread.

## Testing

- Cover approved, denied, malformed, and forged review inputs, plus root rollback and authorization-change scenarios.

GitOrigin-RevId: 27817e1fde9a136de727048c4220d148fcf72f42
2026-08-21 21:22:35 +00:00
Krish Chainani ab8768306f Preserve executor context for MCP stop hooks (#40012)
## What changed

- Scope executor-provided stop-hook calls to the MCP server environment that registered the hook, and reject calls when that environment does not match.
- Forward turn metadata with executor stop-hook requests while retaining the hook call's `threadId` metadata.

## Testing

- Cover environment matching for MCP tool calls and verify executor stop hooks receive their environment ID and request metadata.

GitOrigin-RevId: b544ea3a6ca1f61389fb4597978ed7aac27d529a
2026-08-21 21:17:42 +00:00
Krish Chainani 8b61c50ebe Run allowlisted executor plugin stop hooks (#40009)
## What changed

- Discover inline hooks from executor-provided plugin manifests and accept only
  the bundled Computer Use `Stop` hook for `node_repl.turn_ended`.
- Add the accepted hook to the current step's hook engine with
  executor-scoped provenance.
- Run executor-scoped hooks in the background after regular stop-hook
  decisions, without delaying turn completion, applying control effects, or
  exposing them in hook summaries. Skip them when regular hooks only block
  continuation.

## Testing

- Added coverage for manifest filtering and option preservation.
- Added coverage for background execution, interaction with regular stop
  hooks, and non-blocking turn completion.

GitOrigin-RevId: c0f75f83a23ba8a83bb3a5d5997c4882aacc32d6
2026-08-21 21:01:59 +00:00
felixxia-oai 51d8d12236 Synchronize Git enrichment tests explicitly (#40006)
## What changed

- Track Git enrichment completion with a watch channel and unblock waiters when enrichment finishes or is canceled.
- Let the test sync tool wait for the current turn's enrichment with a bounded timeout.
- Replace polling in metadata tests and wait for enrichment before asserting workspace metadata in the guardian integration test.

GitOrigin-RevId: 8d35f75a0ebfe412674e1e797c6a13c03b1ea373
2026-08-21 20:53:30 +00:00
jif dbe9dac1ae Route escalated commands through synchronous Guardian review (#40005)
## Why

Commands requesting `sandbox_permissions=require_escalated` need a full Guardian review even when they are not marked as retries.

## What changed

- Treat escalated command requests, along with retries, as requiring synchronous Guardian review.
- Bypass extension approval and Guardian V2 shortcuts for these requests.

## Testing

Add an integration test that installs an auto-approving extension and verifies that an escalated command still reaches Guardian and honors its denial.

GitOrigin-RevId: 30eed273460f3c3c6b24d1ce2d29889e34513afd
2026-08-21 20:45:42 +00:00
sayan-oai ad9e8097fd Preserve managed deny-read rules across permission updates (#40004)
## Why

Runtime permission updates must not weaken managed filesystem `deny_read`
requirements.

## What changed

- Retain managed deny-read rules separately and merge them into updated
  permission profiles.
- Reject permission profiles and legacy sandbox policies that conflict with a
  managed denied path.
- Apply the same constraint when `command/exec` handles a request-specific
  sandbox policy.

## Testing

- Cover thread permission updates with managed deny-read requirements.
- Cover `command/exec` enforcement for managed and user-defined denies,
  including conflicting policy and profile overrides.

GitOrigin-RevId: 5e387b9c1bf1650a21753a74a3338bd33df7d0ce
2026-08-21 20:15:23 +00:00
rafael-oai 950dd184a1 Expand browser and computer use requirements (#39995)
## What changed

- Add a top-level `allow_browser_and_computer_use` requirement.
- Support browser requirements for history access, automatic review, persistent approvals, and default or per-origin access policies.
- Support computer-use requirements for persistent approvals, default app access, macOS bundle IDs, Windows AUMIDs, and Windows executables.
- Preserve these settings and their sources when composing layered requirements.

## Testing

- Cover deserialization, empty-value detection, source propagation, and layered TOML merging for the new requirements.

GitOrigin-RevId: d14a10b64465f1c54e9bcc3795dc93031e6b029b
2026-08-21 19:20:09 +00:00
Won Park 56012fafb8 Add Guardian internal session support (#39994)
## What changed

- Add an extension API for spawning host-owned internal sessions and a
  `ThreadReadyInput` lifecycle callback that runs after thread registration.
- Add Guardian reviewer session scaffolding that records the parent thread and
  effective model, plus the under-development `guardian_ext` feature flag.
- Start internal sessions with fresh history while preserving parent lineage,
  shared session controls, and internal-thread visibility rules.
- Scope internal-session prompt cache keys to their source and parent thread,
  and expose `guardian` as an internal session source.

## Testing

- Cover internal-session spawning, parent metadata, history isolation, prompt
  cache keys, and extension spawner argument forwarding.

GitOrigin-RevId: 682dae80397d62bb36247796b447042e760ca364
2026-08-21 19:11:05 +00:00
andrewgu-oai 79b7606803 Keep credentials out of app-server logs (#39993)
## Why

App-server logs can be persisted or included in submitted diagnostics, so credentials used by model providers, authentication refreshes, and attestation requests must not appear in diagnostic output.

## What changed

- Add `RedactedString`, which preserves serialization and string access while replacing debug output with `<redacted>`.
- Use it for model-provider bearer tokens, header and query values, authentication command arguments, and attestation tokens.
- Avoid logging JSON-RPC error payloads and parser or authentication errors that may echo credentials; retain safe context such as error codes and categories.

## Testing

- Add an app-server regression test that exercises provider credentials, refreshed authentication tokens, and attestation tokens, then verifies none appear in persisted SQLite or submitted diagnostic logs.

GitOrigin-RevId: 8c50408adf94d93847658b1320682cf3b637d2cc
2026-08-21 19:04:32 +00:00
sayan-oai f580dd886f Enforce environment network policies for remote execution (#39980)
## Why

`EnvironmentConfig.network_policy` could describe attachment-owned traffic
restrictions, but core rejected every configured policy because execution did
not enforce it.

## What changed

- Resolve each remote environment's network policy for the selected command
  and apply it to the execution-scoped proxy.
- Compose owner rules with controller constraints and saved network decisions
  while preserving inherited domain and Unix-socket denials.
- Keep strict allowlists non-expandable, allow reviewable policies to use
  network approvals, and reject sandbox escalation that would bypass an owner
  policy.
- Reject policies for local execution, disabled managed enforcement, or a
  disabled controller proxy.

## Testing

Added coverage for policy composition, scoped remote proxy behavior, approval
and denial flows, offline execution, and unsupported environment authority.

GitOrigin-RevId: d9331f616df24de6cd13ed68196f0ff7b0ca4dd9
2026-08-21 17:29:41 +00:00
jif d12a7f3fd8 Preserve root user authorization in subagent Guardian reviews (#39975)
## Why

MultiAgent V2 workers can be reviewed after the root conversation has received
new user authorization that is absent from the worker's own transcript. Guardian
reviews need that genuine user context without treating forwarded or
assistant-authored claims as authorization.

## What changed

- Add the bounded root conversation to Guardian prompts for worker reviews.
- Preserve user and assistant roles, and explicitly treat only root user messages
  as authorization evidence.
- Exclude summaries, review artifacts, and assistant commentary from the root
  conversation evidence.
- Apply the same evidence to both core Guardian prompts and Guardian V2
  classification.

## Testing

Add an integration test covering late root-user authorization, forged role text,
forwarded agent claims, and filtered synthetic authorization.

GitOrigin-RevId: 8846de69ba214a88f58f70675798d324e213427c
2026-08-21 17:12:42 +00:00
jif e482cc66ae Keep Guardian reviews isolated from executor MCP servers (#39962)
## Why

Executor-owned MCP configuration can include required servers that are unavailable. Guardian review sessions should not attempt to discover those servers or let them block an approval review.

## What changed

- Skip projecting MCP servers from selected executors when the session source is a Guardian reviewer.
- Preserve executor MCP discovery for regular thread sessions.

## Testing

Added an app-server integration test that configures an unreachable required executor MCP server and verifies that Guardian can still review an escalated command and return its decision.

GitOrigin-RevId: 5b712a3ee04e6896d5ca0cff39a32dd8e0fe0132
2026-08-21 15:44:52 +00:00
jif 3882ced09c Add in-memory shell snapshots to unified exec (#39957)
## What changed

- Add the under-development `shell_snapshot_v2` feature and advertise executor support through environment capabilities.
- Use executor-managed, in-memory snapshots for eligible direct `bash`, `zsh`, and `sh` login commands, while applying the configured shell environment policy and avoiding duplicate automatic startup-file effects.
- Keep file-backed snapshots available for user-shell commands and fall back to the existing execution path when in-memory snapshots are unsupported or inapplicable.

## Testing

- Cover policy filtering, snapshot reuse without snapshot files, automatic startup files, local and remote execution, and legacy user-shell snapshots.

GitOrigin-RevId: 4ad6cdf13824913ac8c393ba38b9844230677579
2026-08-21 15:24:56 +00:00
jif 054588acfe Honor required MCP servers from selected executors (#39952)
## What changed

- Preserve the `required` setting on HTTP MCP servers discovered from a selected executor.
- Cover delayed startup for a required executor MCP server and verify its tool is available on the first turn.

GitOrigin-RevId: df66e12233b8e1aaa7744e29c154157af8dc4776
2026-08-21 15:15:05 +00:00
jif 00a7b888b2 Discover HTTP MCP servers from selected executors (#39941)
## What changed

- Read `mcp_servers` configuration and requirements from each selected remote
  executor and add eligible HTTP servers to the thread's MCP runtime.
- Bind discovered servers to the thread's concrete executor snapshot and apply
  environment MCP policy and requirements. Discovery is best effort, and
  executor-local servers are not treated as required at startup.
- Ignore unsupported stdio servers and HTTP configurations that depend on
  environment-provided headers or header helpers.

## Testing

- Added an app-server integration test covering discovery, authenticated HTTP
  tool invocation, requirements enforcement, and exclusion of stdio servers.

GitOrigin-RevId: 6e1cdcebdbb1cc21a5a2285fbc5617d8d5997182
2026-08-21 14:23:33 +00:00
Tamir Duberstein 748d8ac834 Bound unified exec output delta frames (#39937)
## What changed

- Cap each `ExecCommandOutputDelta` frame at 8 KiB and stop building frames once the per-call delta quota is exhausted.
- Preserve complete UTF-8 scalars across producer chunks while passing malformed or incomplete trailing bytes through without losing them.
- Append all process output to the transcript independently of delta emission.

## Testing

- Cover split multibyte characters, malformed and trailing bytes, frame and event limits, transcript preservation, and interactive PTY output.

GitOrigin-RevId: ef4100e353df01569a8fbae650c70f439920a2fd
2026-08-21 14:04:33 +00:00
Zahan Malkani 536f86e5cc Support attaching to existing realtime calls (#39876)
## Why

Clients that create and negotiate realtime calls themselves need Codex to join
those calls without creating another call or overwriting the client-owned
session configuration.

## What changed

- Add the `existingCall` transport to `thread/realtime/start`, accepting a
  client-provided `callId` and optional `realtimeSessionId`.
- Attach Codex over the call's sideband WebSocket without SDP negotiation or a
  session update, and reconnect that sideband while preserving transcript
  state.
- Default existing calls to realtime v1, support v1 and v3, and reject v2 or
  options that would reconfigure the client-owned session.
- Encode call IDs as a single URL path segment for v3 sideband connections.

## Testing

Add protocol, app-server, core, reconnection, authentication-header, and URL
encoding coverage for existing-call attachments.

GitOrigin-RevId: 2046b70566f98efa55381bb1e461ea3cc68256c5
2026-08-21 06:41:28 +00:00
Brandon Zhang 27c05a52e0 Include context window IDs in response metadata (#39847)
## What changed

- Add `context_window_id` to turn metadata for sampling and compaction requests, using the UUID for the model-visible context window.
- Keep the ID stable for requests in the same context window and advance it after compaction, independently of the existing thread-generation `window_id`.
- Reserve `context_window_id` so client metadata cannot override it.

## Testing

- Cover HTTP and WebSocket metadata, remote compaction, token-budget context alignment, resume behavior, and post-compaction window changes.

GitOrigin-RevId: 55cceea377ad509c8dd96ad668f2b5703ea46a63
2026-08-21 03:31:56 +00:00
Dylan Hurd bd19459358 Ignore project instructions for untrusted projects (#39837)
## What changed

- Skip project-scoped `AGENTS.md` discovery when the active project is untrusted, while preserving user-level instructions.
- Include the active project trust level in the instruction cache key so runtime trust changes reload the applicable instructions.
- Preserve updated project trust metadata during runtime configuration refreshes.

## Testing

Add coverage for initially untrusted projects and for switching between trusted and untrusted states at runtime.

GitOrigin-RevId: 284bf7693943f585013fb5fe4b6aa3309bc50cf3
2026-08-21 02:33:50 +00:00
pmccrary-oai d8ec270183 Rename the history notes extension config option (#39830)
## What changed

Rename the token-budget configuration key and its Rust field from
`use_history_notes_history` to `use_history_notes_extension`. Update the config
schema, extension gating, explicit-setting detection, and tests to use the new
name consistently.

GitOrigin-RevId: b6a514e23f02095fcd1d35dc42eeb7394b800c77
2026-08-21 01:51:48 +00:00
pmccrary-oai daa48072f4 Add history and notes tools for token-budget sessions (#39827)
## Why

Token-budget sessions need a way to recover prior conversation context and preserve working state across context-window transitions.

## What changed

- Add direct-model `history` tools for listing windows and items, reading items, and searching conversation contents.
- Add direct-model `notes` tools for listing, reading, searching, appending, and writing persistent notes.
- Route tool calls through the configured Codex backend with trusted session and agent context, bounded request arguments, and truncation-aware output handling.
- Expose the extension when `features.token_budget.use_history_notes_history` is enabled with an OpenAI provider and Codex backend authentication.

## Testing

- Cover tool registration, configuration changes, provider and authentication requirements, backend request context, encrypted output preservation, and request and response limits.

GitOrigin-RevId: 43b259f01014ba3f30803dd7cd5634942407a9bb
2026-08-21 01:24:52 +00:00
rka-oai f20b63e85c Use Responses compaction for Amazon Bedrock (#39825)
## What changed

- Switch Amazon Bedrock remote compaction to `compaction_trigger` items sent through `/v1/responses`.
- Remove the legacy dedicated compaction protocol from provider capabilities.
- Default remote compaction to unsupported and explicitly enable it for OpenAI, Azure Responses, and Amazon Bedrock providers.

## Testing

- Cover manual and automatic Amazon Bedrock compaction requests and their compacted history.
- Verify remote-compaction capabilities for configured and Amazon Bedrock providers.

GitOrigin-RevId: cf9ba314595367518b37a85f334f0c105ce47e45
2026-08-21 00:53:52 +00:00
chess 67b2c8c6fb Defer legacy filesystem policy projection (#39813)
## Why

Applying session settings rebuilt and compared a legacy filesystem policy even
when the update could not trigger cwd-based policy rebinding.

## What changed

Compute the legacy projection only for cwd changes whose filesystem policy has
a rebindable project-root write entry. Preserve the existing equivalence check
before rederiving the policy.

GitOrigin-RevId: c428a5c54f1b5b1300be406cac374a6fd29bb9a8
2026-08-20 23:36:13 +00:00
chess 969efa5470 Avoid materializing writable-root carveouts for presence checks (#39812)
## What changed

- Add `FileSystemSandboxPolicy::has_writable_roots_with_cwd` to detect an
  effective writable root without constructing its read-only carveouts.
- Use the helper when classifying permission profiles and selecting patch
  rejection reasons.

## Testing

- Verify the presence check matches the existing materialized-root result
  across read-only, unrestricted, external, writable, denied, and unresolved
  special-path policies.

GitOrigin-RevId: 17ee490aa3180e05732133203f68544d9de5268b
2026-08-20 23:26:59 +00:00
rka-oai aead844f64 Handle standalone tool outputs as external context (#39791)
## What changed

- Treat standalone `function_call_output` items without a `call_id` as external context, including items injected into new and forked threads, and mark the thread memory mode polluted when `memories.disable_on_external_context` is enabled.
- Include these outputs in guardian transcripts with their namespaced tool name and a placeholder for non-text content.
- Allow image generation to reuse images from standalone and otherwise unpaired function or custom-tool outputs.

## Testing

- Cover injected thread items, external-context detection, guardian transcript rendering, and recent-image selection.

GitOrigin-RevId: ca4e8c16a4b731e1256a5956908d00f785b4feb1
2026-08-20 19:46:29 +00:00
rka-oai 763787d061 Support standalone named function call outputs (#39782)
## Why

External tool events may need to enter thread history without a preceding function call and therefore do not have a `call_id`.

## What changed

- Allow `function_call_output` items to omit `call_id` and carry optional `name` and `namespace` fields.
- Preserve named standalone outputs during history normalization and agent forks while retaining existing pairing behavior for outputs with a `call_id`.
- Accept, persist, and forward these outputs through `thread/inject_items`, and update the app-server schemas and documentation.

## Testing

- Cover paired and standalone JSON round trips, history normalization, agent forks, and injected thread history.

GitOrigin-RevId: a3258163a7dc93777c7c3023116fe204819bdbb0
2026-08-20 19:19:41 +00:00
tongzhou wang 854cbb2fd4 Make tool-result telemetry limits configurable (#39779)
## What changed

- Add `otel.tool_result.max_bytes`, defaulting to 2048 bytes, to control the
  text included in `codex.tool_result` logs independently of model-visible
  output limits.
- Apply truncation centrally when emitting telemetry, preserve UTF-8
  boundaries, and report whether the logged output was truncated.
- Record typed tool names and namespaces, agent names, and process-local result
  sequence numbers while keeping tool arguments and output out of trace events.

## Testing

- Cover configuration propagation, UTF-8-safe truncation, raw tool-output
  logging, and log/trace event routing and fields.

GitOrigin-RevId: 89ca76e55f1a09b729e7770ce01f2e4d98f9a22d
2026-08-20 19:09:04 +00:00
jif bce5f2fcfc Standardize shell execution on unified exec (#39772)
## What changed

- Use `exec_command` and `write_stdin` as the shell tool surface.
- Treat legacy `default`, `local`, and `shell_command` model metadata as `unified_exec`.
- Remove obsolete shell-selection configuration and runtime paths while preserving the feature and policy gates for zsh fork execution.

GitOrigin-RevId: d743cbe598630d73052f1fecad680c4cde17977d
2026-08-20 18:29:35 +00:00
sayan-oai 5bcd7b0fbc Refresh bundled model definitions (#39770)
## What changed

- Add the hidden Daybreak Blue and Daybreak Red model definitions.
- Refresh model capabilities, instructions, plan availability, and service-tier metadata.
- Configure the auto-review model for Responses Lite and code-mode tools, and update Guardian request tests for the resulting `additional_tools` and developer-message layout.

## Testing

- Update Guardian request snapshots and assertions for initial and follow-up reviews.
- Verify MCP approval routing against the developer-message form of the Guardian prompt.

GitOrigin-RevId: 6680e9abebdcbc43224a81348591e26f8422f3ec
2026-08-20 18:17:28 +00:00
william-openai 39073ca3a7 Include suggestion IDs in plugin install metadata (#39765)
## What changed

- Add `suggestion_id` to plugin install elicitation metadata, using the same
  value as the elicitation request ID and plugin-install analytics event.
- Leave connector install metadata unchanged by omitting the field.

## Testing

- Cover plugin and connector metadata serialization.
- Verify that remote plugin install metadata and analytics carry matching
  suggestion IDs.

GitOrigin-RevId: 05dd1d210e2a82857094f09d27ecba79ee460b7e
2026-08-20 17:54:42 +00:00
victor-openai 097825f75a Add app-server MCP event streaming (#39761)
## What changed

- Add experimental `mcpServer/event/stream/start` and `mcpServer/event/stream/stop` requests for hosted apps, plus `mcpServer/event/stream/notification` forwarding.
- Scope subscriptions to the owning app-server connection and subscribed thread, enforce unique IDs and a per-connection limit, and clean them up when the thread is unsubscribed or the connection closes.
- Wait for the MCP active notification before completing startup, retry streams that close immediately, and terminate them when authentication or hosted runtime ownership changes.

## Testing

- Add an app-server integration test covering activation, event forwarding, duplicate subscription rejection, and explicit cancellation.

GitOrigin-RevId: 8a6fc1615adfc5af7e67def3b824fa5909ab3e7b
2026-08-20 17:49:35 +00:00
jif 8a40095ea3 Standardize shell execution on unified exec (#39757)
## What changed

- Remove the legacy `shell_command` handler and runtime, leaving `exec_command`
  and `write_stdin` as the shell execution tools.
- Treat legacy `shell_command` model metadata as `unified_exec`, and normalize
  legacy user opt-outs so they do not disable command execution. Managed feature
  requirements and `shell_tool` can still disable it.
- Preserve shell approvals, sandboxing, zsh-fork support, and output truncation
  through the unified execution path.

## Testing

- Cover legacy configuration and model-metadata compatibility.
- Exercise unified shell execution, approvals, truncation, and `apply_patch`
  serialization across the app-server and core test suites.

GitOrigin-RevId: 5c2fd6164fc3519cdae4944cb9db276b8467311c
2026-08-20 17:46:05 +00:00
jif d0cc662b8c Cache shell snapshots in the exec server (#39756)
## What changed

- Add the `shellSnapshotV2` executor capability and an optional shell snapshot request to `ExecParams`.
- Capture and restore Unix shell state and profile exports from an in-memory, attachment-scoped cache for `bash`, `zsh`, and `sh`.
- Apply environment policies, runtime `PATH` entries, sandbox context, and live managed-proxy settings when preparing restored commands.
- Bound snapshot size, capture time, scope length, and cache capacity, and fall back to the original command when capture fails.

## Testing

- Cover local, remote, TTY, sandboxed, and supported-shell execution, plus environment filtering, proxy handling, in-memory reuse, and capture failure fallback.

GitOrigin-RevId: 624f747972c249c88c6f10f42cf0af97b75b5541
2026-08-20 17:39:06 +00:00
joeflorencio-openai ce950dcf26 Add managed developer instructions to requirements (#39755)
## What changed

- Add `additional_developer_instructions` to managed requirements and expose it through `configRequirements/read` as `additionalDeveloperInstructions`, independently of ordinary developer instructions.
- Include the managed instructions in model context, emit explicit replacement or removal messages when requirements change, and preserve the current value across compaction, resume, and agent forks without duplication.
- Reject managed instructions whose rendered context exceeds 10,000 estimated tokens.

## Testing

- Cover requirements layering and API serialization, context updates and removal, size validation, repeated model requests, compaction and resume, rollout migration, and agent forks.

GitOrigin-RevId: bc0b70fb7988944c2f68176dff55f5ed61eb46c8
2026-08-20 17:36:22 +00:00
Sean Huang 85a1b0e33d Expose uncompiled permission profile selection (#39752)
## What changed

- Add `resolve_permission_profile_selection` to select the effective profile using configured defaults, managed requirements, and allowlists.
- Return the merged configured and managed profile catalog without compiling platform-specific paths, leaving path interpretation to the executor.

## Testing

- Verify that a managed default overrides a configured default while preserving Windows-style paths.
- Verify that an allowlisted but undefined profile is rejected.

GitOrigin-RevId: 0172193d48f0346522f931d3b806054895d199c0
2026-08-20 17:25:14 +00:00
cgst-oai a26d50852a Require filesystem backends to implement directory walks (#39749)
## What changed

- Make `ExecutorFileSystem::walk` a required backend operation instead of
  providing a fallback built from directory reads and metadata requests.
- Implement bounded local walks on a blocking task with cancellation, symlink
  cycle detection, deterministic ordering, error collection, and response-size
  limits.
- Have remote filesystems use the server's walk operation directly.

## Testing

- Cover local and remote handling of invalid roots and limits, directory
  symlinks, non-UTF-8 names, cancellation, sandbox contexts, and response
  budgets.

GitOrigin-RevId: 7499bf05080c3f9965a5eb7ffd593de604d62c2a
2026-08-20 17:16:52 +00:00
sayan-oai 2e1f18e0db Refresh resumed thread capability roots from executors (#39746)
## Why

Resumed threads can carry a persisted capability-root location that no longer
matches the location reported by the attached executor.

## What changed

- Refresh matching thread-owned capability roots from the live environment while
  preserving persisted roots when the executor reports none.
- Use the same merge behavior for capability-root inspection and per-step MCP
  resolution, without changing owner-configured root precedence.

## Testing

Extend the remote-environment resume test to cover both an executor-provided
replacement root and the persisted-root fallback.

GitOrigin-RevId: 0c66195b2406f431a2e156e00d5225a8d22d18b2
2026-08-20 17:12:15 +00:00
rhan-oai 4f38432d87 Use model-specific auto-review outcome instructions (#39741)
## What changed

- Add `rejection_instructions` and `timeout_instructions` to catalog-provided auto-review messages.
- Use the acting model's instructions for denied and timed-out reviews across tool approvals, shell escalation, and MCP elicitation responses.
- Fall back to the existing instructions only when a catalog value is absent, while preserving explicit empty-string overrides.

## Testing

- Cover catalog overrides, legacy fallbacks, empty values, and separation between acting-model and reviewer-model messages.

GitOrigin-RevId: c5b2c2dbdaefd45d1d658651dd1abaeb6d8c93da
2026-08-20 16:52:16 +00:00
jif 9894a14c81 Track multi-agent v2 spawn calls in analytics (#39722)
## What changed

- Emit started and completed collaboration tool events for multi-agent v2
  `spawn_agent` calls, including failed invocations.
- Record execution duration and successful child-agent configuration metadata
  without including the spawn prompt.
- Deduplicate collaboration and subagent activity items by call ID when
  calculating per-turn subagent tool counts.

## Testing

- Extend the app-server multi-agent v2 integration test to cover successful
  and failed spawn telemetry, duration, prompt omission, and turn counts.

GitOrigin-RevId: c1c2fb9cc2a1a9b0f63ef68509ad36d11a92d1e5
2026-08-20 14:33:55 +00:00
rafael-oai 1674b0a130 Expose managed policy for browser settings imports (#39720)
## What changed

- Add `in_app_browser.allow_external_browser_settings_import` to managed requirements, preserving explicit Boolean values through layered composition while leaving an omitted value unset.
- Return the policy as `inAppBrowser.allowExternalBrowserSettingsImport` from `configRequirements/read` and include it in the generated protocol schemas.
- Keep the import policy independent from the in-app browser feature flag and agent Browser Use requirements.

## Testing

- Cover parsing, managed-layer precedence, invalid values, user and session override resistance, and app-server response serialization.

GitOrigin-RevId: efa2621d2b1cf503f1bee2505d9914cb4fb7221d
2026-08-20 14:13:56 +00:00
Tamir Duberstein 4a942885c8 Reduce unified exec output buffer allocations (#39712)
## What changed

- Accept borrowed byte slices when appending output chunks, avoiding clones before
  forwarding those chunks to streaming consumers.
- Merge drained head-tail buffers by reusing their owned storage while preserving
  the retained prefix, latest suffix, and omitted-byte count.
- Make unified exec output handles capacity-generic so bounded-output tests can
  exercise repeated drains with small buffers.

## Testing

- Update head-tail buffer and process manager tests to cover chunk retention,
  omission accounting, and repeated output drains with the new APIs.

GitOrigin-RevId: caa9813d3c9d9921cff28888263a212def381640
2026-08-20 13:17:20 +00:00
Ben Romano 312b62ac95 Trace MCP runtime refresh coordination (#39667)
## What changed

Add named tracing spans for MCP dirty-state refreshes and time spent waiting
to acquire the refresh semaphore.

GitOrigin-RevId: 1dbb3572d2061581beb83b3de098d6ec91791104
2026-08-20 08:36:00 +00:00
pakrym-oai e3e5ad2847 Harden unsandboxed patch filesystem access (#39659)
## Why

An `apply_patch` path can be replaced with a symlink after verification, allowing an unsandboxed patch operation to reach a different file than the one that was approved.

## What changed

- Add `follow_symlinks` options to executor filesystem reads, writes, metadata lookups, directory creation, and removal, including the corresponding `followSymlinks` protocol fields.
- Implement no-follow filesystem operations on Unix and Windows that reject links in any path component and restrict file access to regular files.
- Run `apply_patch` with symlink traversal disabled when an otherwise-required sandbox is bypassed, while retaining the existing follow-symlink default for standalone callers.

## Testing

- Cover leaf and ancestor symlinks across patch add, update, delete, and move operations, including a path swap after verification.
- Exercise local and remote no-follow filesystem behavior, concurrent directory creation, special-file rejection, and Windows reparse points.

GitOrigin-RevId: 43fd479084891493ce13564fbd894b98f329c6dd
2026-08-20 08:10:08 +00:00
Ankush Gupta 4e1a772a7d Let Guardian V2 satisfy required model reviews (#39658)
## What changed

- Allow the Guardian V2 approval monitor to handle reviews for models that require automatic review.
- Preserve full Guardian review when Guardian V2 is disabled by configuration or managed requirements.
- Cover required-model routing for low- and high-risk actions, plus both Guardian V2 disable paths.

GitOrigin-RevId: d1201cc3dbd7c4a07de633388b14f17ada9feb39
2026-08-20 08:07:06 +00:00
Adam Perry @ OpenAI 7ece061767 Enforce filesystem permissions when loading AGENTS.md (#39653)
## Why

Project instructions must respect the selected environment's filesystem read
permissions. Tightening those permissions for a later turn must not allow
previously cached instructions to reach the model.

## What changed

- Apply each environment's filesystem sandbox while discovering and reading
  `AGENTS.md` files.
- Fail thread or turn setup when sandboxing blocks a discovered instruction
  file, while allowing a restricted project with no instructions to start.
- Clear cached instructions before refresh and include the Windows sandbox
  level in the cache key.
- Ignore inaccessible ancestor marker probes so readable instructions in the
  selected working directory can still load.

## Testing

Added coverage for restricted projects, denied instruction files, cache
invalidation after permissions tighten, and unreadable ancestor markers.

GitOrigin-RevId: 6ea1a27b9c873a6260e4f87d42ae1317a6a4ae4b
2026-08-20 07:48:46 +00:00
Dylan Hurd 8aaf839774 Exercise restricted-token sandboxing in cyber policy tests (#39646)
## What changed

Run both branches of the heuristically safe command policy test with the
Windows sandbox level set to `RestrictedToken`.

GitOrigin-RevId: ae5d754ec4a51d73a8e5f9d844f94a91f795fb72
2026-08-20 07:37:52 +00:00
Anton Panasenko af0e82c562 Enforce managed residency for model providers (#39645)
## Why

Model provider configuration could override the residency header required by
`enforce_residency`, so provider-backed requests did not reliably honor the
managed setting.

## What changed

- Apply the managed residency header after building a provider, making it
  authoritative for model requests and model discovery.
- Warn when a provider configures the residency header through
  `http_headers` or `env_http_headers`, while preserving the original provider
  configuration and unrelated headers.
- Cover HTTP and WebSocket requests, model discovery, and case-insensitive
  header detection with targeted tests.

GitOrigin-RevId: 461ef8989d5bd06ffdea694aee4265e392e07590
2026-08-20 07:33:27 +00:00
evanz-oai 663da53823 Sanitize developer context in full-history agent forks (#39641)
## Why

Developer messages can contain both inherited agent policy and unrelated context. Full-history forks need to replace the parent policy without dropping the unrelated content that shares its message.

## What changed

- Filter fork-specific developer instructions by content item instead of excluding an entire developer message.
- Remove inherited multi-agent mode instructions before applying the child agent's current mode.
- Preserve unrelated content items and discard developer messages only when filtering leaves them empty.

## Testing

Extend agent-control and subagent notification tests to cover compound compacted messages, multi-agent mode instructions, and proactive-to-explicit mode transitions.

GitOrigin-RevId: 028834e237d8636c13b17a724574bc8eb09a55ba
2026-08-20 07:29:43 +00:00
Sean Huang 7ea7b29369 Expose permission profile resolution in the core API (#39632)
## What changed

- Add `Config::resolve_permission_profile` to resolve named profiles from the
  effective configuration and managed requirements.
- Make active-profile network proxy lookup public and re-export
  `CodexThreadSettingsOverrides` from `codex-core-api`.
- Cover configured and managed profile inheritance, duplicate-name rejection,
  and preservation of managed network settings.

GitOrigin-RevId: ffb5f54a932ef1a4b624677b5f051ecdcd1134b8
2026-08-20 07:11:46 +00:00
jif 942af8447b Retire the untrusted approval policy (#39630)
## What changed

- Remove `untrusted` from the CLI, configuration schema, and MCP tool interface. Explicit `approval_policy = "untrusted"` settings now fail with an actionable error.
- Remove the known-safe command allowlist. Projects marked untrusted now request approval for every command unless an explicit exec policy rule allows it.
- Keep command parsing conservative by treating in-place `sed` forms as mutating and ignoring unrecognized commands when recording memory usage.

## Testing

- Cover rejection of the retired configuration value and approval requests for commands in untrusted projects.

GitOrigin-RevId: d6bf425edddfffbb325eee6acf383434af5fd33b
2026-08-20 07:03:02 +00:00