## Summary
This keeps the apply_patch approval tests focused on approval behavior
instead of macOS sandboxed filesystem helper startup.
The changed cases still force patch approval with `UnlessTrusted`, but
use `DangerFullAccess` after approval so the patch write is direct and
cheap. Workspace-write and sandbox-helper behavior remain covered by the
filesystem and apply_patch sandbox tests.
## Why
Prepare network approval scoping to distinguish execution environments
without changing behavior yet.
## What changed
- Add optional environment IDs to network policy requests.
- Add optional network environment IDs to exec and sandbox request
structs.
- Thread default None values through existing construction points.
- Fix stale constructor call sites that caused the CI compile failures.
## Not included
- Per-environment proxy listeners.
- Network approval cache or prompt behavior changes.
- Ambiguous request attribution handling.
Those behavior changes moved to stacked follow-up #28899.
## Validation
- just fmt
- CI will run tests and clippy
## What
This PR defines the structured configuration contract for shared rollout
token budgets (across ALL agent threads under 1 rollout).
```toml
[features.rollout_budget]
enabled = true
limit_tokens = 100000
reminder_interval_tokens = 10000
sampling_token_weight = 1.0
prefill_token_weight = 0.1
```
The reminder interval defaults to 10% of the rollout limit. Sampling and
prefill weights default to `1.0`.
## Scope
This PR only defines and validates configuration. It does not track
usage, inject reminders, or stop a rollout. Accounting and reminders are
implemented in the stacked follow-up #28494.
The existing `token_budget` feature remains unchanged. `rollout_budget`
has its own feature key and configuration type.
## Tests
The config test verifies that the structured fields resolve into
`RolloutBudgetConfig` and do not enable the existing `token_budget`
feature.
Local checks:
- `just write-config-schema`
- `just test -p codex-core load_config_resolves_rollout_budget`
- `cargo check -p codex-thread-manager-sample`
- `git diff --check`
The full workspace test suite was not run locally.
## What changed
- retain the parsed plugin manifest namespace on loaded plugins
- carry that namespace through `PluginSkillRoot` and `SkillRoot`
- use the provided namespace when qualifying plugin skill names
- include the namespace in the skills cache key
## Why
Plugin loading has already parsed `plugin.json`, but skill parsing
currently walks every `SKILL.md` ancestor and probes/reads the manifest
again to reconstruct the same namespace. Passing the parsed namespace
removes those repeated filesystem calls, which are particularly costly
on remote filesystems.
Context:
https://openai.slack.com/archives/C0ARA9GF5D4/p1781639496496439?thread_ts=1781202444.891669&cid=C0ARA9GF5D4
## Impact
Plugin skill names remain unchanged. A regression test uses a
deliberately different on-disk manifest name to verify that plugin roots
use the provided parsed namespace.
## Validation
- `just test -p codex-core-skills -p codex-core-plugins -p codex-plugin
-p codex-utils-plugins` (352 passed)
- `just fix -p codex-core-skills -p codex-core-plugins -p codex-plugin
-p codex-utils-plugins`
- `just fmt`
## What changed
- promote plugin config loading to an info-level `plugins_for_config`
span
- promote skill config loading to an info-level `skills_for_config` span
- attach stable OpenTelemetry names to both spans
## Why
`session_init.plugin_skill_warmup` currently combines plugin loading and
skill loading, which makes cold-start traces unable to identify which
phase dominates. These child spans preserve the existing aggregate while
making the two costs independently visible.
Context:
https://openai.slack.com/archives/C0ARA9GF5D4/p1781639496496439?thread_ts=1781202444.891669&cid=C0ARA9GF5D4
## Impact
This is observability-only. It does not change plugin or skill loading
behavior.
## Validation
- `just test -p codex-core-skills -p codex-core-plugins` (347 passed)
- `just fmt`
## Why
App-server must report command events containing foreign-platform paths
without changing existing client or rollout path-string formats.
## What changed
- retain `PathUri` through exec command begin/end events
- convert cwd values to `LegacyAppPathString` at the app-server
compatibility boundary
- drop command actions with foreign paths and log them
- serialize rollout-trace cwd values using their inferred native path
representation
- restore Wine coverage for retained Windows cwd values and successful
completion
## Summary
Allow plugin manifests to declare `skills` as either a single path
string or an array of path strings in the core plugin loader.
## Why
Some plugin packages need to expose skills from more than one directory.
Before this change, `plugin.json` only accepted a single string for
`skills`, so manifests like this were ignored as an invalid `skills`
shape:
```json
{
"skills": ["./skills/abc", "./skills/edk"]
}
```
This keeps the existing single-string form working while adding support
for the list form. The final scope is intentionally limited to the core
plugin manifest/load path for `skills`; `apps`, file-backed
`mcpServers`, and the bundled plugin-creator assets are unchanged in
this PR.
## What changed
- Parse `skills` as either a string or an array of strings in
`plugin.json`.
- Store resolved skill paths as a list in `PluginManifestPaths`.
- Load manifest-declared skill roots in addition to the default
`./skills` root.
- Deduplicate exact duplicate skill roots before loading.
- Rely on existing skill-loader dedupe by canonical `SKILL.md` path for
overlapping roots such as `./skills` plus `./skills/abc`.
- Update plugin manifest tests to cover:
- single string `skills`
- list of string `skills`
- duplicate skill roots
- `./skills` as a manifest path
- explicit child roots like `./skills/abc` and `./skills/edk`
- overlapping-root dedupe
## Validation
- `just test -p codex-plugin`
- `just test -p codex-core-plugins`
- `just test -p codex-mcp-extension`
- `git diff --check`
## Why
Som tools, such as history and notes, must remain top-level when MCP
deferral is enabled while staying unavailable through code-mode `exec`.
## What changed
- Added `features.code_mode.direct_only_tool_namespaces`.
- Classified matching MCP tools as `DirectModelOnly`.
- Kept those tools top-level in `code_mode_only`.
- Excluded them from `tool_search` deferral and the nested `exec`
surface.
- Updated the generated config schema.
## Validation
- `code_mode_only_exposes_direct_model_only_mcp_namespaces`
- `load_config_resolves_code_mode_config`
## Summary
- add a provider API that supplies a fresh signed WebSocket URL for each
remote exec-server connection
- refresh the signed URL after disconnects and retry once when a
handshake returns `401 Unauthorized`
- allow `EnvironmentManager` consumers to register remote environments
backed by the URL provider
## Tests
- `just test -p codex-exec-server -E
'test(remote_websocket_client_refreshes_url_after_unauthorized_handshake)
| test(remote_websocket_client_refreshes_url_after_disconnect)'` — 2
passed
- `cargo check -p codex-core-api` — passed
- `just fix -p codex-exec-server` — passed
- `just fix -p codex-core-api` — no test targets; no-op
- `just fmt` — passed
- `just test -p codex-exec-server` — 187 passed; 32 unrelated macOS
sandbox tests could not invoke nested `sandbox-exec` (`Operation not
permitted`)
## Why
Frontend realtime voice continuity needs to replay a tiny
previous-session overlap as actual conversation items, including
assistant text. The app-server `thread/realtime/appendText` API already
carries a role through to the Rust realtime websocket layer, but the
shared role enum only accepted `user` and `developer`.
## What Changed
- Added `assistant` to `ConversationTextRole` and regenerated the
app-server schema/type fixtures.
- Added `output_text` as a realtime conversation content type.
- Updated realtime websocket item creation so assistant appendText emits
`content: [{ type: "output_text", text }]`, while user and developer
continue to emit `input_text`.
- Updated app-server docs and tests to cover assistant appendText
alongside the existing developer role behavior.
## Validation
- `just write-app-server-schema`
- `just fmt` (first sandboxed attempt failed because `uv` could not
access `~/.cache/uv`; reran with filesystem access and passed)
- `just test -p codex-api` passed: 126/126
- `just test -p codex-app-server-protocol` passed: 239/239, including
generated JSON/TypeScript fixture checks
- `just test -p codex-app-server` was started locally but stopped per
request after unrelated local sandbox/Seatbelt failures (`sandbox-exec:
sandbox_apply: Operation not permitted`) and one missing local `codex`
binary failure; CI should be faster and more authoritative for the full
suite.
## What
Built on the realtime speech-control plumbing merged in #27917.
- Add optional `codexResponseHandoffPrefix` to `thread/realtime/start`.
- Apply that prefix only to automatic V1 commentary sent through
`conversation.handoff.append`; final answers remain unprefixed.
- Add opt-in `clientManagedHandoffs`. When true, core suppresses
automatic response handoffs and completion output so delivery is
controlled by explicit client append APIs.
- Preserve existing automatic behavior by default.
`codexResponsesAsItems: true` continues to select item routing when
client-managed mode is disabled.
## Why
Voice clients need two delivery policies: automatic background context
with silent commentary instructions and fully client-owned handoffs.
Phase-aware prefixing keeps routine commentary silent without
suppressing the final answer, while client-managed mode lets an app
decide exactly which updates to append.
## Validation
- `just fmt`
- `cargo test -p codex-app-server-protocol
serialize_thread_realtime_start`
- `RUST_MIN_STACK=16777216 cargo test -p codex-core --test all
conversation_handoff_persists_across_item_done_until_turn_complete`
- `RUST_MIN_STACK=16777216 cargo test -p codex-app-server --test all
webrtc_v1_client_managed_handoffs_disable_automatic_output`
- `RUST_MIN_STACK=16777216 cargo test -p codex-app-server --test all
webrtc_v1_final_automatic_handoff_omits_silent_prefix`
- `cargo build -p codex-cli --bin codex`
- Local Codex Apps compatibility check: 43 focused webview tests passed,
and a live voice session routed through the source-built app-server.
The explicit `RUST_MIN_STACK` avoids a macOS Tokio test-worker stack
overflow seen with the default test environment.
## Why
A durable realtime voice orchestrator can reconnect and resume through
multiple fresh `Session` instances. Realtime handoffs were using the
Session-local `auto-compact-N` counter as their turn identity, but that
counter restarts at zero for every resumed Session. The durable thread
could therefore accumulate duplicate turn IDs, violating the uniqueness
assumptions made by app-server and web clients. In Codex Apps, a new
delegated response stream could be attached to an older turn with the
same ID, placing live output higher in history and putting turn-scoped
actions at risk.
Persisted rollout and reconstructed model-context order were already
correct because raw response items remain append-only and chronological.
This change restores unique identity for reconstructed and live turn
surfaces.
## What changed
- Generate a UUIDv7 specifically for each realtime-routed delegation.
- Leave the existing `auto-compact-N` identity path unchanged for actual
internal auto-compaction turns.
- Extend the inbound realtime handoff integration test to require a UUID
turn ID from `turn/started`.
## Verification
- `just test -p codex-core inbound_handoff_request_starts_turn`
- `just fix -p codex-core`
- `just fmt`
## 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
Codex can run a remote exec server through the Noise relay, but the
normal
environment-manager path could not establish an
environment-registry-backed
harness connection. Signed rendezvous URLs and harness authorizations
are
short-lived, so reconnects must fetch a fresh bundle instead of
retaining
stale connection credentials. A stalled registry request must also fail
within
the regular remote connection deadline, without exposing these
credentials in
debug logs.
Issue: N/A (internal environment-service integration).
## What Changed
- Add environment-manager configuration for a registry-backed Noise
rendezvous
environment.
- Request a fresh bundle from
`/cloud/environment/{environment_id}/connect` for every physical harness
connection, using the existing 10-second remote connection timeout.
- Share the Environment Registry register, connect, and validate wire
payloads
through `codex-exec-server` and `codex-core-api`.
- Redact the signed rendezvous URL and harness authorization from the
public
connect response's `Debug` output.
- Add focused coverage for registry bundle retrieval, stalled requests,
and
credential redaction.
## Why
`PathUri::join` should not depend on the app-server compatibility
wrapper `LegacyAppPathString` to parse native paths. Native path parsing
belongs to the URI abstraction that it constructs.
## What
Move platform-independent native path parsing into the root `PathUri`
module. `PathUri::join` and `LegacyAppPathString` now share the
crate-private `PathUri::from_absolute_native_path` constructor.
## Summary
- allow local marketplace `source.path: "."` and `source.path: "./"` to
resolve to the marketplace root
- keep `""` invalid and preserve rejection of non-root paths without
`./` plus non-normal/traversal paths
- add focused regression coverage for repo-root plugin layouts and
rejected local paths
## Tests
- `RUSTUP_TOOLCHAIN=stable just fmt`
- `RUSTUP_TOOLCHAIN=stable just test -p codex-core-plugins`
- `RUSTUP_TOOLCHAIN=stable just fix -p codex-core-plugins`
Note: plain pinned-toolchain `just fmt` was blocked locally by a rustup
`clippy` component conflict, so validation used the working stable 1.95
toolchain fallback.
## Why
Services that proxy the exec-server environment registry endpoints need
to deserialize and forward the same Noise registration and harness-key
validation payloads. Those wire models currently live as private,
serialize-only structs in `exec-server`, which forces consumers to
duplicate the contract.
## What changed
- Add owned serde models for registration and harness-key validation
requests and responses.
- Use those models in the existing exec-server registry client.
- Re-export the models from `codex-exec-server` and `codex-core-api`.
- Keep the harness authorization request free of a derived `Debug`
implementation so it is not accidentally logged.
## Testing
- Focused exec-server registration and harness-key validation tests: 2
passed.
- `cargo check -p codex-core-api`
The full `codex-exec-server` suite compiled and ran 254 tests: 222
passed, while 32 existing filesystem sandbox tests could not run under
the nested macOS sandbox (`sandbox_apply: Operation not permitted`).
Co-authored-by: Codex <noreply@openai.com>
## Summary
- Track plugin install failures through the unified
`codex_plugin_install_failed` event for local installs, remote install
preflight failures, bundle failures, and remote catalog/backend
failures.
- Send classified `error_type` values in plugin install failure
analytics instead of raw error strings.
- Stop sending raw external-agent import errors in analytics while
preserving raw failure details in app-facing import
notifications/history.
- Keep raw plugin/migration diagnostics in `tracing::warn!` logs.
- Keep remote failure plugin names as the existing local placeholder
(`unknown`) and remove the extra telemetry plugin-name override.
- Change `ExternalAgentConfigImportParams.source` from a generated enum
to `string | null`, with legacy `claudeCode` / `claudeCowork` inputs
normalized to existing analytics values.
## Testing
## Why
It should be possible for app-server to handle "foreign" OS paths in
unified_exec working directories, allowing e.g. a Linux app-server to
run processes on e.g. a Windows exec-server.
## What
Convert the core unified_exec cwd values to use `PathUri`.
Adds fallible path conversion in several places to try to minimize the
scope of this change. The only time this change suppresses errors from
converting `PathUri` to an `AbsolutePathBuf` is when the turn is
configured with no sandboxing at all to allow us to make progress
testing without sandboxing.
Future changes to apply_patch and sandboxing will clean up these error
paths.
A tool's cwd is resolved from joining a model-provided workdir to the
environment's cwd. When using `AbsolutePathBuf::join()`, an
absolute-path workdir would overwrite the environment's cwd and we would
resolve permissions/sandboxing against the model-provided path. This
change extends `PathUri::join()` to also treat an absolute rhs as an
override of the base/lhs.
This also removes some coverage from the remove_env_windows tests until
a follow-up converts foreign paths in command exec events correctly.
## Breaking Changes
When using `AbsolutePathBuf::join()` for workdir resolution, we ended up
resolving tilde-prefixed paths against the app-server's `$HOME`, e.g.
`~/foo/bar` becomes `/home/anp/foo/bar`. It's difficult to do this with
`PathUri` joining, so after offline discussion this PR no longer
implements it.
A quick check of some power users' rollouts suggests that models don't
actually generate home-prefixed absolute working directories for their
spawns, so this shouldn't have any real blast radius.
This mechanically extracts the existing TUI plugin catalog and detail
popup rendering from `chatwidget/plugins.rs` into a new
`chatwidget/plugin_catalog.rs` module. `plugins.rs` now keeps the
stateful plugin workflow and orchestration, while `plugin_catalog.rs`
owns the presentation-heavy catalog/detail popup construction and its
pure helpers. The goal is to keep `plugins.rs` focused before later
plugin sharing work adds more catalog behavior.
- Moves existing catalog/detail popup builders and related pure helpers
into `plugin_catalog.rs`
- Leaves plugin fetch/state/key handling in `plugins.rs`
- Adds only minimal sibling-module visibility/import wiring
- Intentionally makes no product behavior or UI changes beyond the code
move
## Summary
- Revert #28655, restoring the thread `recencyAt` behavior introduced by
#27910.
- Move `threads_recency_at` to migration 0039 so it no longer collides
with `external_agent_config_imports` at version 0038.
- Repair databases that already applied the recency migration as version
38 by moving the matching migration-history row to version 39 before
SQLx validation. The current version-38 migration can then apply
normally.
## Validation
- `just test -p codex-state
migrations::tests::repairs_recency_migration_that_was_applied_as_version_38`
- `just test -p codex-state -p codex-rollout -p codex-thread-store -p
codex-app-server-protocol -p codex-tui`: 3,439 passed; six TUI tests
could not open the machine's existing read-only incident database at
`~/.codex/sqlite/state_5.sqlite`.
- `just fix -p codex-state`
- `just fmt`
- Verified that state migration versions are unique.
## Stack
This is PR 1 of the simplified HAI single-run-task stack:
- [#19047](https://github.com/openai/codex/pull/19047) Agent Identity
assertion and task-registration primitives, including the shared
run-task helper used by existing Agent Identity JWT auth.
- [#19049](https://github.com/openai/codex/pull/19049)
Disabled-by-default ChatGPT auth opt-in that provisions/reuses persisted
Agent Identity runtime auth and its single run task.
- [#19051](https://github.com/openai/codex/pull/19051) Run-scoped
provider auth that uses one backend-owned task id for first-party
inference and compaction requests.
[#19054](https://github.com/openai/codex/pull/19054) collapsed out of
the active stack because the simplified design no longer needs a
separate background/control-plane task helper.
## Summary
The simplified POC shape is one backend-owned task per Agent Identity
run. This PR makes the first layer match that final shape directly
instead of introducing task targets, caller-owned external task refs, or
intermediate wrappers that later PRs would need to undo.
What changed:
- keeps the `AgentAssertion` wire payload as `agent_runtime_id`,
`task_id`, `timestamp`, and `signature`
- exposes `register_agent_task` as the single task-registration helper
for both existing Agent Identity JWT auth and the ChatGPT-registration
path added later in the stack
- makes task registration send only the signed registration timestamp;
the backend owns the returned opaque task id
- removes the unused target/task-kind/external-task-ref surfaces from
`codex-agent-identity`
- keeps Agent Identity JWT JWKS lookup separate from agent/task
registration URL derivation
- updates Agent Identity JWT auth to register one run task during auth
construction and share that task across cloned auth handles
This PR intentionally does not enable ChatGPT-derived Agent Identity.
That opt-in and config gate are added in the next PR.
## Testing
- `just test -p codex-agent-identity`
## Why
Command approval cache keys included the command and working directory,
but not the execution environment. An approval for `/workspace` locally
could therefore be reused for the same command and path on an executor.
## What changed
- Include the selected environment ID in shell and unified-exec approval
cache keys.
- Carry that ID through the normal command approval request so clients
can show which environment is being approved.
- Expose the environment through app-server as a required nullable
`environmentId` and show it in the inline TUI approval prompt.
- Keep older recorded approval events compatible when the environment is
absent.
For example, `echo ok` in local `/workspace` and `echo ok` in executor
`/workspace` now produce different approval keys and separate prompts.
## Scope
This PR does not change network approvals, Guardian review actions, MCP
elicitation, full-screen TUI rendering, or environment-ID validation.
Remote `shell_command` execution itself remains in #28722; this PR only
makes its approval key environment-aware.
## Why
The community marketplace audit found many skill frontmatter parse
failures where values were intended as prose, but were not valid YAML.
Common examples include unquoted scalar values with `: `, such as
`description: Build for AWS: ECS` or `argument-hint: <duration: e.g.
7d>`, and flow-looking values such as `tags: [next,@supabase/ssr]`.
`serde_yaml` does not expose a permissive mode for this. The parser
fails before unknown frontmatter fields can be ignored, so a
compatibility repair has to happen before retrying YAML parsing.
## What changed
Skill frontmatter loading still uses `serde_yaml` as the primary parser.
If that parse fails, the loader performs a line-oriented repair of
scalar frontmatter field values, then retries parsing.
The fallback now:
- applies to any frontmatter mapping field, not just `description` /
`short-description`
- quotes unquoted scalar values that contain a YAML colon separator such
as `: `
- quotes invalid flow-looking scalar values that start with `[`, `{`,
`@`, or backtick
- preserves already quoted values
- skips `|` / `>` block scalar bodies so multiline descriptions are not
rewritten
- returns the original YAML error if the repaired frontmatter still
cannot parse
## Examples
This previously failed because the second `: ` was parsed as YAML
structure:
```yaml
description: AWS deployment patterns: ECS Fargate, Lambda, and S3
```
The fallback now parses it as if it had been written explicitly as:
```yaml
description: 'AWS deployment patterns: ECS Fargate, Lambda, and S3'
```
The same repair now applies to ignored frontmatter fields that still
need to be valid YAML for the parser to get through the document:
```yaml
argument-hint: <duration: e.g. 7d, 2w>
tags: [next,@supabase/ssr]
```
Valid YAML multiline descriptions continue to work through normal
parsing without repair:
```yaml
description: |-
Build for AWS: ECS
and Lambda
```
## Validation
- Added loader coverage for unquoted `description` values containing `:
`.
- Added loader coverage for unquoted `metadata.short-description` values
containing `: ` and an apostrophe.
- Added loader coverage for unrecognized frontmatter fields that need
quoting, including `argument-hint` and `tags`.
- Added block-scalar coverage to ensure multiline description bodies are
preserved while other fields are repaired.
- `just test -p codex-core-skills` (106 passed)
- `just fix -p codex-core-skills`
## Why
This is the final PR in the Windows fs-helper sandbox stack and contains
the actual bug fix.
The exec-server filesystem helper is a direct-spawn path: it asks
`SandboxManager` for a `SandboxExecRequest`, then launches the returned
argv itself. That works on macOS and Linux because the transformed argv
is already a self-contained sandbox wrapper. On Windows, the transformed
request carried `WindowsRestrictedToken` metadata, but the direct-spawn
fs-helper runner still launched the helper argv directly.
That means Windows filesystem built-ins backed by the fs-helper could
run with the parent Codex process permissions instead of the configured
Windows sandbox. This PR makes the direct-spawn transform produce a
self-contained Windows wrapper argv before fs-helper launches it.
## What Changed
- Added `SandboxManager::transform_for_direct_spawn()` for callers that
launch the returned argv themselves.
- Wrapped Windows restricted-token direct-spawn requests with `codex.exe
--run-as-windows-sandbox` and then marked the outer request as
unsandboxed, matching the macOS/Linux wrapper argv shape.
- Updated `exec-server/src/fs_sandbox.rs` to use the direct-spawn
transform for fs-helper launches.
- Materialized the inner `codex.exe --codex-run-as-fs-helper` executable
into `.sandbox-bin` so the sandboxed user can run it.
- Carried runtime workspace roots through `FileSystemSandboxContext` as
`PathUri` values so `:workspace_roots` policies resolve correctly
without sending native client paths over exec-server JSON.
- Preserved wrapper setup identity environment needed by Windows sandbox
setup without changing the serialized inner helper environment.
## Verification
- `just bazel-lock-update`
- `just bazel-lock-check`
- `just test -p codex-sandboxing transform_for_direct_spawn_windows`
- `just test -p codex-exec-server fs_sandbox::tests`
- `just fix -p codex-windows-sandbox -p codex-sandboxing -p
codex-exec-server -p codex-core -p codex-file-system`
Local note: `just fmt` completed Rust formatting, but this workstation
still fails the non-Rust formatter phases because uv cannot open its
cache and the local buildifier/dotslash path is missing.
## 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
The app server is long-lived, but its shared model cache otherwise
refreshes only when a caller needs it. Once the five-minute cache
expires, starting a thread or calling `model/list` can wait for
`/models` on the request path.
Refresh the cache in the background before it expires so foreground
callers normally use fresh local state.
## What changed
- Start an app-server worker that refreshes models immediately and then
every three minutes using the existing models-manager API.
- Hold only a weak reference to the models manager between refreshes, so
the worker does not extend its lifetime.
- Stop scheduling refreshes when the app-server lifecycle handle is shut
down or dropped. A refresh already in progress is allowed to finish.
- Adjust affected app-server test fixtures to distinguish the background
`/models` probe from the connection they are testing.
The existing models-manager cache, refresh strategies, auth handling,
ETag behavior, and concurrency semantics are unchanged.
## Testing
-
`models_refresh_worker::tests::refreshes_immediately_periodically_and_stops_when_dropped`
-
`suite::v2::remote_control::listen_off_honors_persisted_remote_control_enable`
-
`suite::v2::attestation::attestation_generate_round_trip_adds_header_to_responses_websocket_handshake`
## Summary
This keeps inter-agent communication on the existing raw response item
path and adds a join key for MAv2 tool calls.
MAv2 `spawn_agent`, `send_message`, and `followup_task` now stamp the
originating tool call id into `ResponseItemMetadata.source_call_id` on
the raw `ResponseItem::AgentMessage`. App-server clients can join that
raw item back to the existing tool/activity event by call id, while
using the raw agent message's existing sender, receiver, and content
fields.
No new app-server `ThreadItem` or notification type is added.
## Tests
- `just fmt`
- `just write-app-server-schema`
- `just test -p codex-protocol`
- `just test -p codex-app-server-protocol`
- `just test -p codex-core
multi_agent_v2_spawn_returns_path_and_send_message_accepts_relative_path`
- `just test -p codex-core
multi_agent_v2_followup_task_completion_notifies_parent_on_every_turn`
- `just fix -p codex-protocol`
- `just fix -p codex-app-server-protocol`
- `just fix -p codex-core`
## Why
PR #28512 retries a failed session recovery every 100 ms. Every Noise
recovery attempt first asks the environment registry for a fresh
connection bundle, even when the eventual failure comes from the
WebSocket or initialize handshake. During an outage, that could make
each disconnected client call the registry about 250 times during the
25-second recovery window.
## What changes
All retryable Noise recovery failures now use a separate backoff
schedule:
```text
base: 500 ms -> 1 s -> 2 s -> 4 s -> 5 s maximum
actual: 500-750 ms, 1-1.5 s, 2-3 s, 4-6 s, 5-7.5 s
```
The extra 0-50% is deterministic per-session jitter so disconnected
clients do not retry together. Direct WebSocket recovery keeps the
existing 100 ms retry because it does not re-enter the registry.
Supersedes #28288 (closed).
## Why
A short WebSocket interruption currently ends every client-side process
handle, even though exec-server keeps the server session and its
processes alive for a short time.
This is especially visible for executor-backed stdio MCP servers: a
temporary connection loss becomes a permanent `Transport closed` error.
The server already has the information needed to resume the session, but
the client opens a fresh session instead of using it.
This change reconnects below the process and MCP layers. Existing
process handles stay valid, missed output is recovered, and the same
server-side processes continue running.
## State machine
One logical `ExecServerClient` stays alive while its underlying RPC
connection changes generations.
```text
transport closes
+------------------------------------------------+
| v
+-------------+ +-------------+
| Connected | | Recovering |
+-------------+ +-------------+
^ |
| session resumed, processes caught up | retryable error
+------------------------------------------------+ loops until deadline
|
| deadline or permanent error
v
+-------------+
| Failed |
+-------------+
```
### `Connected`
- New RPC calls use the current connection.
- Process notifications are published in sequence order.
- A disconnect only starts recovery if it came from the current
connection generation. Late events from older generations cannot replace
the active connection.
### `Recovering`
- New calls wait instead of choosing a half-connected RPC client.
- Existing process handles, wake subscriptions, and event subscriptions
stay open.
- Streaming HTTP response bodies fail immediately because their byte
streams cannot be resumed safely.
- Recovery first waits for process starts that were already in flight. A
start whose result became ambiguous is cleaned up after reconnection
instead of being silently adopted.
- The client reconnects with the learned `session_id`. The server may
briefly report that the old connection is still attached, so that error
is retried until the detach finishes.
- The notification consumer starts before the resume handshake
completes. This prevents a busy process from filling the notification
queue and blocking the initialize response.
- Before installing the new connection, the client catches up every
recoverable process with `process/read`.
### `Failed`
- Recovery stops after 25 seconds or after a permanent error.
- Waiting calls are released with one stable disconnect error.
- Existing process sessions receive a terminal failure instead of
waiting forever.
## Recovering process events
Output, exit, and close events share one sequence. During normal
operation, the client buffers early events until every lower sequence
has been published.
After reconnection, the client reads each process starting after its
last published sequence:
1. Retained output chunks are inserted by sequence number.
2. Exit and close state are reconstructed in their sequence positions.
3. Events already received as live notifications are ignored as
duplicates.
4. Newly contiguous events are published in order.
5. If the server no longer retains enough output to fill a sequence gap,
only that process is terminated and failed. The recovered connection
remains usable for other processes.
The server reports its full next event sequence for unbounded reads,
including exit and close events. Closed processes remain readable for
the same 30-second window used to retain detached sessions.
## Other details
- Detached server sessions are retained for 30 seconds, leaving margin
around the client's 25-second recovery deadline.
- Session attach and detach update the active notification sender under
the same attachment lock, so an old connection cannot clear a newly
attached sender.
- A dedicated error code distinguishes the temporary "session is still
attached" race from permanent initialization errors.
- Process starts are identity-checked on both client and server. Cleanup
from an older start cannot remove a newer process that reused the same
ID.
- Mutating requests that were already in flight when the transport
closed are not replayed, because the client cannot know whether the
server applied them. Requests started after recovery is known wait for
the replacement connection.
- We assume the server/client version stays in sync (on the before/after
this PR)
## User impact
Long-running commands and stdio MCP servers can survive a temporary
exec-server WebSocket interruption without changing process IDs or
losing output produced during the outage.
## Why
#27920 stopped persisting image-generation items unless their status was
`completed`, preventing failed standalone extension items with empty
results from being saved. Built-in image generation can instead emit a
terminal `response.output_item.done` containing a complete base64 PNG
while the item status remains `generating`. In that case, app-server
emits no `savedPath`, so Codex Apps can render the inline image but
cannot expose a file artifact.
## What changed
- Persist image-generation items whenever `result` contains image data.
Failed terminal items still have empty results and remain unpersisted.
- Update the existing built-in image-generation integration test to
cover a terminal `generating` item and verify both `saved_path` and the
written PNG bytes.
## Validation
- Confirmed with a raw built-in websocket trace: the image progressed
through `in_progress`, `generating`, and `partial_image`, then emitted
one `response.output_item.done` with `status: "generating"` and a
complete PNG result.
- `just test -p codex-core builtin_image_generation_call_persisted` is
currently blocked before test execution by a pre-existing compile error
in `thread-store/src/thread_metadata_sync.rs:171`.
## Why
`TurnContext` had accumulated dead fields and cached projections of
values already owned by its per-turn `Config` or `ModelInfo`. Keeping
both copies made ownership unclear and allowed artificial split-brain
states, such as a compatibility hash differing from the model metadata
it came from.
`Prompt` similarly carried a write-only personality after personality
selection had already been materialized into its base instructions.
This makes the canonical owner explicit: configuration-backed values
come from `config`, model-derived values come from `model_info`, and
prompts contain only data consumed by request construction.
## What changed
- Remove the unused `ghost_snapshot`, `codex_self_exe`, and
`thread_source` fields.
- Remove duplicate `comp_hash`, `truncation_policy`, `features`,
`shell_environment_policy`, `codex_linux_sandbox_exe`, `compact_prompt`,
and `tool_mode` fields.
- Read those values directly from `TurnContext::config` or
`TurnContext::model_info` at their consumers.
- Remove the write-only `Prompt::personality` field and its constructor
assignments.
- Preserve review-turn inheritance of the parent turn's shell policy,
Linux sandbox executable, and compact prompt through the review config.
## Testing
- `cargo check -p codex-core --tests`
## Why
Revert #27910 to remove the newly introduced thread `recencyAt`
persistence and API behavior from `main`.
## What changed
This reverts commit `fac3158c2a783095768076489815f361fa9b0db4`,
including the state migration, thread-store propagation, app-server API
surface, generated schemas, and related tests.
## Validation
Not run before opening; relying on CI for the initial fast signal.
## Summary
Code mode has two separate truncation points: the nested tool result
returned to JavaScript and the code-mode output later recorded for the
model. These tests now verify those behaviors independently.
- Report whether `result.output` was truncated before printing it.
- Verify omitted or sufficiently large nested limits produce `Variable
truncated: False`, while allowing the printed value to be truncated
downstream.
- Verify an explicit nested limit produces `Variable truncated: True`
when the command output exceeds it.
- Use a token-policy model fixture so downstream truncation is visible
as `…N tokens truncated…`.
- Align the explicit nested-truncation expectation with the warning
header.
This PR changes test coverage only; runtime truncation behavior is
unchanged.
## Validation
- `env -u CODEX_SANDBOX_NETWORK_DISABLED RUST_MIN_STACK=8388608 cargo
test -p codex-core --test all code_mode_exec -- --nocapture` (8 passed)
A code-mode cell is a single JavaScript execution that can produce
output, call tools, wait for asynchronous work, resume, or be
terminated. This PR extracts the existing per-cell run loop into a
dedicated actor that owns the cell’s lifecycle state. It is primarily an
ownership change rather than a new lifecycle contract: existing behavior
now has one clear implementation boundary.
### Architecture
The session service remains responsible for session-wide concerns:
allocating cell IDs, storing shared values, creating cells, and routing
requests to them.
Once a cell is created, its execution state belongs to its actor.
Callers interact with the actor through a handle. The actor receives two
kinds of input: runtime events and control requests.
A single event loop serializes these inputs and applies the lifecycle
rules. It tracks the current observer—the caller waiting for an
update—along with accumulated output, outstanding callbacks, runtime
state, yield deadlines, and termination progress. Observation,
termination, completion, and cleanup therefore have one consistent
owner.
When the runtime has no immediately runnable work and is waiting only on
timers or tool results, the actor can return accumulated output and
information about outstanding tool calls while keeping the cell
available to resume. On completion or termination, it performs the
appropriate callback cleanup before publishing the final result and
removing the cell from the session.
A small host interface connects the actor to session-owned facilities
such as tool dispatch, notifications, stored values, and final cell
removal, keeping those responsibilities outside the actor itself.
### Why
Previously, cell lifecycle state and coordination lived alongside
session management. The actor boundary makes each cell a self-contained
state machine with a single writer, while the service becomes a registry
and adapter around it.
This makes lifecycle behavior easier to reason about and test in
isolation. It also establishes a clean boundary for later changing where
cells run or how they communicate without recreating their lifecycle
rules.
## Summary
This fixes plugin manifest parsing for MCP servers declared as an object
directly in `plugin.json`.
Before this change, Codex modeled `mcpServers` as only a string path,
for example:
```json
{
"name": "counter-sample",
"version": "1.1.1",
"mcpServers": "./.mcp.json"
}
```
Some migrated plugins instead provide the server map directly in the
manifest:
```json
{
"name": "counter-sample",
"version": "1.1.1",
"description": "Plugin that declares MCP servers in the manifest",
"mcpServers": {
"counter": {
"type": "http",
"url": "https://sample.example/counter/mcp"
}
}
}
```
That object form previously failed during install/load with an error
like:
```text
failed to parse plugin manifest: invalid type: map, expected a string
```
## What changed
- Add a manifest representation for `mcpServers` as either
`Path(Resource)` or `Object(map)`.
- Parse `plugin.json` `mcpServers` as either a string path or an object.
- Route object-valued MCP server maps through the existing plugin MCP
config parser instead of adding a second parser.
- Apply existing per-plugin MCP server policy to object-valued MCP
servers the same way as file-backed MCP servers.
- Include object-valued MCP server names in plugin telemetry/capability
metadata.
- Support object-valued MCP config for executor plugins without
requiring a `.mcp.json` filesystem read.
- Update the bundled plugin-creator validator and `plugin-json-spec.md`
so generated-plugin validation accepts the same object-valued shape.
## Compatibility
Existing plugin manifests that use `"mcpServers": "./.mcp.json"`
continue to work. Plugins can now also use the object shape shown above.
## Tests
Added coverage for the new manifest attribute shape at the install,
normal load, telemetry, and executor-provider layers:
- `install_accepts_manifest_mcp_server_objects`
- `load_plugins_loads_manifest_mcp_server_objects`
- `plugin_telemetry_metadata_uses_manifest_mcp_server_objects`
- `reads_manifest_object_config_without_executor_file_system_access`
Also smoke-tested the plugin-creator validator against both supported
forms:
- `mcpServers` as a direct object in `plugin.json`
- `mcpServers` as `"./.mcp.json"` with a companion `.mcp.json`
## Validation
- `just test -p codex-plugin`
- `just test -p codex-core-plugins`
- `just test -p codex-mcp-extension`
- `just bazel-lock-update`
- `just bazel-lock-check`
- `just fmt`
- `git diff --check`
- Focused rename/object-form rerun: `just test -p codex-core-plugins
manager::tests::load_plugins_loads_manifest_mcp_server_objects
manager::tests::plugin_telemetry_metadata_uses_manifest_mcp_server_objects
store::tests::install_accepts_manifest_mcp_server_objects`
- Focused executor rerun: `just test -p codex-mcp-extension
executor_plugin::provider::tests::reads_manifest_object_config_without_executor_file_system_access`
- `python3
codex-rs/skills/src/assets/samples/plugin-creator/scripts/validate_plugin.py
/private/tmp/codex-validator-object`
- `python3
codex-rs/skills/src/assets/samples/plugin-creator/scripts/validate_plugin.py
/private/tmp/codex-validator-path`
## Why
A `codex-thread-store` test fixture still constructs
`ResponseItem::FunctionCallOutput` without its required `metadata`
field, preventing the crate's test targets from compiling on `main`.
## What changed
- Set the fixture's response-item metadata to `None`.
## Testing
- `cargo check -p codex-thread-store --tests`
## Why
#28152 jumped the gun on moving the rollout format to store URIs, and
would likely break compat with some features that don't go through the
same types as the core logic.
## What
Make `TurnContextItem.cwd` an `AbsolutePathBuf` again, remove test added
for `PathUri` serialization in rollouts. Also drops a bunch of error
paths that are no longer needed.
## Summary
- Treat the remote global plugin catalog as active only when
`remote_plugin` is enabled and the current auth uses the Codex backend.
- Skip the local OpenAI curated marketplace for remote-enabled ChatGPT
users while preserving configured marketplaces.
- Keep the local curated marketplace for API-key users, unauthenticated
fallback, and ChatGPT users with `remote_plugin` disabled.
- Apply the same effective-remote gate to the remote
installed-marketplace cache.
## Root cause
The tool-suggestion discovery path unconditionally included the local
OpenAI curated marketplace. For remote-enabled ChatGPT users, that made
remote discovery additive: Codex parsed every local curated
`plugin.json` before also loading the remote catalog.
## Validation
- `just fmt`
- `cargo build -p codex-cli --bin codex`
- Targeted auth/feature matrix tests pass, including API-key auth with
`remote_plugin` enabled.
- Manual CLI validation confirmed:
- ChatGPT + remote off includes local curated.
- ChatGPT + remote on excludes local curated.
- API-key auth keeps local curated when remote is enabled.
- `just test -p codex-core-plugins`: 235 passed; one unrelated existing
marketplace test failed because it loaded the developer's home
marketplace configuration.
## Summary
Add a server-owned `recencyAt` timestamp and `recency_at` thread-list
sort key for product recency ordering while preserving the existing
meaning of `updatedAt` as the latest persisted thread mutation.
This is the server-side alternative to #27697. Rather than narrowing
`updatedAt`, clients can sort the sidebar by `recency_at` and continue
treating `updatedAt` as mutation time.
Paired Codex Apps PR:
[openai/openai#1024599](https://github.com/openai/openai/pull/1024599)
## Contract
- `recencyAt` initializes when a thread is created.
- A turn start advances `recencyAt` monotonically.
- Commentary, agent output, tool results, token/accounting updates, turn
completion, archive, unarchive, resume, and generic metadata writes do
not advance it.
- `updatedAt` retains its existing behavior and continues to advance for
persisted thread mutations.
- Current servers populate `recencyAt`; the response field is optional
in generated TypeScript so clients connected to older servers can fall
back to `updatedAt`.
- Filesystem-only fallback uses existing updated/mtime ordering when
SQLite is unavailable.
## Persistence and compatibility
Migration 0038 adds second- and millisecond-precision recency columns,
backfills them from the existing updated timestamp, creates list
indexes, and includes an insert trigger so older binaries writing to a
migrated database seed recency without causing later mutations to
advance it.
Generic metadata upserts preserve existing recency values. Turn-start
updates use a dedicated monotonic touch, and process-local allocation
keeps millisecond cursor values unique. State DB list, search, read,
filtered-list repair, rollout fallback propagation, and app-server
conversions all carry the new field.
## API
`Thread` responses include:
```ts
recencyAt?: number
```
`thread/list` and `thread/search` accept:
```json
{ "sortKey": "recency_at" }
```
Generated TypeScript and JSON schemas are included.
## Validation
- `just test -p codex-state` — 146 passed
- `just test -p codex-rollout` — 69 passed
- `just test -p codex-thread-store` — 81 passed
- `just test -p codex-app-server-protocol` — 231 passed
- Focused app-server list ordering, response mapping, archive/unarchive,
and resume lifecycle tests passed
- Scoped `just fix` for state, rollout, thread-store,
app-server-protocol, and app-server
- `just fmt`
- `git diff --check`
- Independent correctness, simplicity, elegance, security, and
test-quality reviews; actionable ordering, lifecycle, query-projection,
and timestamp-uniqueness findings were addressed
## Summary
Introduces the default-off `respect_system_proxy` feature flag used to
gate first-class system PAC/proxy support for Codex-owned native
clients.
With the feature disabled or absent, behavior remains unchanged. This PR
establishes the configuration and managed-requirement surface; proxy
discovery and request routing are implemented by follow-up PRs.
## Configuration
User configuration uses the standard boolean feature form:
```toml
[features]
respect_system_proxy = true
```
Managed feature requirements use the corresponding boolean key. The
effective runtime configuration is exposed as a boolean and defaults to
`false`.
## Implementation
- Registers `respect_system_proxy` as an under-development, default-off
feature.
- Resolves user configuration and managed feature requirements into
`Config.respect_system_proxy`.
- Provides bootstrap resolution for startup paths that must evaluate the
feature before full configuration loading completes.
- Uses the standard feature CLI and config-editing behavior.
- Excludes `features.respect_system_proxy` from project-local
configuration.
- Updates the generated configuration schema.
## End-user behavior
- No networking behavior changes when the feature is absent or disabled.
- Enabling the feature makes the boolean available to the native
proxy-routing implementation in follow-up PRs.
- Repository-local configuration cannot enable the feature.
## Test coverage
Covers scalar configuration and CLI override resolution, managed
requirement constraints, bootstrap resolution, and project-local
filtering.
## Why
Model-visible `<environment_context>` should match the environment of
the executor, not of the app server.
Stacked on #28146.
## What
- Keep selected environment cwd values as `PathUri` while building
environment context.
- Render cwd text using the path convention represented by the URI, with
the canonical URI as a fallback.
- Preserve compatibility with legacy `TurnContextItem.cwd` values when
reconstructing and diffing context.
- Extend the Wine-backed remote Windows test to assert that the model
sees `powershell` and `C:\windows`.
Summary\n- Await endpoint recommendation selection while constructing
each authenticated turn, removing the first-turn cache race.\n- Snapshot
and filter endpoint candidates once per turn, then use that same set for
the bounded contextual user fragment, tool exposure, and exact install
validation.\n- Keep recommendation selection ephemeral: do not persist
recommendation state in or gate resumed threads on prior context.\n-
Hide the legacy list tool in endpoint mode and preserve legacy discovery
unchanged when the endpoint is disabled or unavailable.\n- Keep remote
plugin and connector app identities out of model-visible context and
attach them only to Codex-owned elicitation metadata.\n\nStack\n- 3/4,
based on #28400.\n- Endpoint client and cache: #28399.\n- Generalized
suggestion presentation: #28400.\n- Install-schema follow-up:
#28403.\n\nValidation\n- \n- \n- \n- \n- Full : 2,649 passed and 88
environment-dependent tests failed because this sandbox cannot write ,
nest Seatbelt, or locate auxiliary test binaries.
Summary
- Add list-backed and developer-context presentations for plugin
suggestion candidates.
- Let tool planning, install validation, and request-tool copy follow
the selected presentation.
- Keep every production caller on the existing list-backed presentation,
preserving the current list tool, request schema, connector behavior,
and model-visible copy.
- Leave developer-context presentation latent until the final PR in the
stack.
Stack
- 2/3, based on #28399.
- Follow-up: #27704 activates endpoint recommendations.
Validation
- `just test -p codex-core request_plugin_install`
- `just test -p codex-core spec_plan`
- `just fix -p codex-core`
- `just fmt`
- `git diff --check`
Summary
- Add authenticated parsing for `/ps/plugins/suggested?scope=GLOBAL`,
including remote plugin and connector app identities.
- Validate, deduplicate, sort, and cap endpoint candidates before
caching them by backend and account identity.
- Deduplicate concurrent cache misses and warm recommendations from the
existing remote-installed-plugin refresh path used at startup and after
account changes.
- Keep endpoint results model-invisible in this PR; failures and
responses without `enabled: true` resolve to legacy mode.
Stack
- 1/3. Follow-up: #28400 generalizes plugin suggestion presentation
without activating endpoint recommendations.
- Final activation: #27704.
Validation
- `just test -p codex-core-plugins recommended_plugins`
- `just fix -p codex-core-plugins`
- `just fmt`
- `git diff --check`
## 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`
jifandGitHub
rka-oaiandGitHub
Matthew ZengandGitHub
Adam Perry @ OpenAIandGitHub
charlesgong-openaiandGitHub
Won ParkandGitHub
Anton PanasenkoandGitHub
guinness-oaiandGitHub
jiayuhuang-openaiandGitHub
pakrym-oaiandGitHub
Owen LinandGitHub
Casey ChowandGitHub
Konstantine KahadzeandGitHub
canvrno-oaiandGitHub
Jeremy RoseandGitHub
AdrianandGitHub
iceweasel-oaiandGitHub
Alex ZamoshchinandGitHub
Ahmed IbrahimandGitHub
Channing CongerandGitHub
xl-openaiandGitHub
Alex DaleyandGitHub