## Why
Compaction analytics adds retained image count and compaction summary
output tokens for v1.5 specifically.
## What changed
- Add nullable `retained_image_count` and `compaction_summary_tokens`
fields to `codex_compaction_event`.
- Populate them only for `responses_compaction_v2`: retained images come
from the retained v2 compacted history, and summary tokens come from
`response.completed.token_usage.output_tokens`.
- Leave local and legacy remote compaction events as `null` for these
detail fields.
## Verification
- `just fmt`
- `just fix -p codex-core`
- `just test -p codex-core
build_v2_compacted_history_counts_retained_input_images`
- `git diff --check`
## Summary
- Keep the existing `x-codex-window-id` HTTP header unchanged.
- Also send the same window ID in Responses `client_metadata`, allowing
supported backend paths to surface it as
`x-client-meta-x-codex-window-id`.
- Cover normal HTTP Responses and remote compaction v2 requests without
changing window generation or compaction behavior.
## Why
In the `2026-06-06T23` production hour, all 28,729 HTTP compaction
requests had `window_id` in `x-codex-turn-metadata`, but only 73
retained the direct `x-codex-window-id` header. The request-body
`client_metadata` path is already used for installation ID and is
preserved through supported Responses API paths.
This is additive metadata only. It does not change the direct header,
request count, model input, compaction routing, window generation, or
user response behavior.
Legacy `/v1/responses/compact` is intentionally unchanged. Its current
server-side `CompressBody` schema does not accept `client_metadata` and
rejects unknown fields, so supporting that path requires a backend
schema change before the Codex client can safely send this field.
## Validation
- Current head: `219baef3c`, rebased onto `origin/main` at `26d932983`.
- The post-rebase diff remains limited to the original five files (`22`
insertions, `6` deletions); the legacy experiment remains fully
reverted.
- `just test -p codex-core
responses_stream_includes_subagent_header_on_review`: passed; validates
normal HTTP Responses metadata.
- `just test -p codex-core
remote_compact_v2_reuses_compaction_trigger_for_followups`: passed;
validates remote compaction v2.
- `just test -p codex-core
remote_manual_compact_chatgpt_auth_reuses_service_tier_and_prompt_cache_key`:
passed; validates that legacy compact keeps its accepted payload shape.
- `just test -p codex-core
remote_manual_compact_api_auth_omits_service_tier_and_reuses_prompt_cache_key`:
passed; validates the legacy API-key payload as well.
- `just fmt`: passed; an unrelated root `justfile` rewrite produced by
the formatter was discarded.
- `git diff --check origin/main...HEAD`: passed.
The focused server pytest could not start in the local monorepo
environment because test setup is missing the `dotenv` module. Server
source and tests explicitly show that `CompressBody` omits
`client_metadata` and `/v1/responses/compact` returns HTTP 400 for
unknown body fields.
## Why
Remote-control app-server sessions can reconnect every 5-7 seconds when
the shared transport-event queue fills. The queue's consumer handled
`ConnectionClosed` by awaiting all in-flight RPCs for the disconnected
connection. A stuck RPC therefore blocked processing of replacement
connection and initialize events until remote-control forwarding hit its
five-second timeout and reconnected again.
Related issue: N/A (internal remote-control incident investigation).
## What Changed
- Split fast RPC admission closure from draining:
`ConnectionRpcGate::close()` rejects queued and future RPCs, while
`shutdown()` continues waiting for RPCs that already started.
- Close a disconnected connection's RPC gate before spawning the
existing RPC drain and resource cleanup in a tracked background task, so
the transport-event consumer remains available without waiting for
active RPCs.
- Reap completed cleanup tasks during normal operation, drain them
during graceful shutdown, and abort them during forced shutdown.
- Add regression coverage for closing with an active RPC, rejecting
post-close requests without polling them, and preserving the existing
shutdown wait behavior.
## Verification
`just test -p codex-app-server --lib connection_rpc_gate` passes all 6
tests, including the new close-versus-drain regression coverage.
## Summary
- Restore separate release symbol archives for macOS, Linux, and Windows
binaries.
- Build release binaries with `line-tables-only` debuginfo instead of
full debuginfo.
- Strip Unix distribution binaries after extracting symbols, preserve
Windows PDBs, and keep symbol archives available to the release job.
- Strip the packaged Linux `bwrap` binary before hashing it so the
embedded digest matches the distributed bytes.
## Root cause
The first symbol-artifact implementation enabled
`CARGO_PROFILE_RELEASE_DEBUG=full`. In the June 2 release runs, macOS
ARM primary builds reached the 90-minute timeout while still inside
`Cargo build`. After the symbol changes were reverted, the same primary
build completed in about 22 minutes. The archive step itself completed
in tens of seconds when reached.
Rust's `line-tables-only` debuginfo level preserves function names and
source locations for symbolication without emitting the heavier variable
and type information from full debuginfo.
## Validation
- Ran `just fmt` from `codex-rs`.
- Ran `just test-github-scripts` from the repository root: 23 tests
passed.
- Ran `bash -n` and `shellcheck` on
`.github/scripts/archive-release-symbols-and-strip-binaries.sh`.
- Parsed both modified workflows as YAML and ran `git diff --check`.
- Built a macOS release smoke binary with `line-tables-only`, archived
its dSYM through the restored script, stripped the production binary,
and verified that `atos` resolves `symbol_smoke_function` to
`main.rs:2`.
- Ran Linux archive-script control-flow coverage with stubbed `objcopy`
and `strip` commands.
- Ran Windows PDB archive staging coverage and verified
underscore-emitted Rust PDB names are staged under shipped hyphenated
binary names.
## Follow-up
The release workflow only runs for tags or manual dispatches, so CI
cannot dry-run the full release matrix on this PR. The next release run
will verify runner time and memory behavior under `line-tables-only`.
## Summary
- add contains_external_context() to tool output so other tools can be
opted out of influencing memory when disable_on_external_context=true
- Classify standalone web-search output as external context (to match
behavior as hosted web search)
- Verify with integration test
## Why
Multi-agent v2 residency is intended to keep only the threads that need
to be live. The existing rollout resume path still walked persisted open
descendants and reopened the entire descendant tree when resuming a v2
root, which turns resume into an eager reload of work that should stay
unloaded until it is explicitly needed.
The interrupted-agent path has a related residency issue. Interrupted
agents remain open by design, so an idle interrupted resident should be
eligible for eviction just like an idle completed or errored resident.
Otherwise a resident set full of interrupted agents can consume every v2
slot and block later spawns or reloads with `AgentLimitReached`.
## What Changed
- Return early from `resume_agent_from_rollout` after resuming a v2
thread so persisted v2 descendants are not reopened eagerly.
- Treat idle `Interrupted` v2 residents as unloadable in the LRU
residency path.
- Add focused coverage for v2 root resume leaving descendants unloaded
and for eviction of an idle interrupted v2 resident when a new slot is
needed.
## Verification
Added targeted `codex-core` tests covering:
- v2 root resume with persisted descendants, verifying only the root is
loaded after resume.
- residency eviction of an idle interrupted v2 agent when the resident
set is full.
## Why
`close_agent` is the wrong model-facing name for the v2 operation after
the residency changes. V2 agents remain reusable by task name, and
residency/unloading owns capacity management; the exposed tool should
describe the action it actually performs: interrupt the target agent's
current turn without making the agent unavailable for future messages or
follow-up tasks.
## What changed
- Rename the multi-agent v2 tool from `close_agent` to
`interrupt_agent`.
- Keep the v1 `close_agent` surface unchanged.
- Update the v2 handler to send `Op::Interrupt`, keep interrupted agents
registered, and reject root/self targets with interrupt-specific errors.
- Route interrupt delivery through the existing dead-thread cleanup path
so stale resident entries do not keep consuming capacity.
- Update tool planning and handler tests for the new v2 surface and
semantics.
## Verification
Added focused coverage in:
- `core/src/tools/spec_plan_tests.rs`
- `core/src/tools/handlers/multi_agents_tests.rs`
## Why
Multi-Agent V2 concurrency should count active non-root turns, not
resident or durable agent threads. The limit is intentionally best
effort: admission checks are synchronous, but concurrent successful
checks may overshoot slightly.
## What changed
- Keep one root-derived execution limit on the shared `AgentControl`.
- Count active V2 subagent turns with an RAII guard owned by
`RunningTask`.
- Check capacity before spawning or starting an idle agent, including
direct app-server `turn/start` submissions.
- Preserve queued delivery for agents that are already running.
- Exempt automatic idle continuations so `/goal` work is not dropped
when capacity is temporarily full.
- Keep root and V1 turns outside this limiter.
## Test coverage
- `execution_guards_count_active_v2_subagent_turns`
- `execution_guards_ignore_root_and_v1_turns`
- `v2_nested_spawn_checks_shared_active_execution_capacity`
## Summary
- ignore RUSTSEC-2026-0173 in cargo-deny and cargo-audit config
- document that proc-macro-error2 is pulled in transitively via
i18n-embed-fl/age/codex-secrets
- leave the ignore temporary until codex-secrets moves off age or age
drops i18n-embed-fl
## Validation
- just fmt
- cargo deny check --hide-inclusion-graph
## Why
Multi-agent v2 treats agents as durable logical agents, not just live
entries in `ThreadManager`. After the reload-on-delivery change, a v2
agent can be addressed even if its thread is not currently loaded.
This PR adds the next layer: loaded v2 subagents can be paged out of
`ThreadManager` when the session has too many resident agents. That
keeps residency separate from logical identity and prepares the stack
for making v2 concurrency count active execution instead of existing
agents.
## What Changed
- Add an `AgentControl`-scoped LRU for resident v2 subagents.
- Reserve residency before spawning or reloading a v2 subagent.
- If resident capacity is full, unload the least-recently-used idle v2
subagent from `ThreadManager`.
- Keep `ThreadManager` as a primitive loaded-thread store; it does not
own the LRU policy.
- Keep unloaded agents registered and durable so they can be reloaded by
the delivery path.
- Preserve the existing v2 cap semantics by using the derived non-root
v2 cap for residency.
Eviction is intentionally conservative. A thread is unloadable only when
it is a v2 subagent, has completed or errored, has no active turn, and
has no pending mailbox work. Before removal, the rollout is materialized
and flushed.
## Assumptions And Non-Goals
- PR #26623 provides the reload-on-delivery path for unloaded v2 agents.
- `ThreadManager` membership means loaded/resident, not logical agent
existence.
- `AgentRegistry` remains the logical identity/metadata source for v2
agents that may be unloaded.
- `list_agents` remains a recent/resident view for now.
- This does not change active execution concurrency; that is the next
PR.
- This does not change `close_agent` semantics.
- This does not change or remove `resume_agent`.
- This does not add a new residency config knob.
## Stack
1. V2 durable lookup and reload on delivery (#26623) - reload unloaded
v2 agents before delivering follow-up/input.
2. V2 residency LRU (this PR) - unload idle resident v2 agents from
`ThreadManager` when resident capacity is full.
3. V2 active-execution concurrency - count running non-root v2 turns
instead of logical agents.
4. V2 close/interrupt semantics - make v2 close interrupt the current
turn without deleting durable identity.
5. V2 resume cleanup - remove the manual resume surface for v2 while
keeping internal reload support.
## Validation
- Added focused coverage for the residency LRU eviction path.
- Local clippy/check/tests were not run; CI will cover them.
## What
- Consume plaintext `output` from standalone search while retaining
optional `encrypted_output` parsing.
- Expose `web.run` to code mode and return search output to nested
JavaScript calls.
- Cover direct and code-mode standalone search paths with integration
tests.
## Why
`/v1/alpha/search` now returns plaintext output, which code mode needs
to consume standalone search results.
## Test plan
- `just test -p codex-api`
- `just test -p codex-web-search-extension`
- `just test -p codex-core code_mode_can_call_standalone_web_search`
- `just test -p codex-app-server
standalone_web_search_round_trips_output`
## Why
MCP startup failures from spawned subagents were rendered as global
notifications, so a child thread's failure could pollute the visible
parent transcript. Routing the notification to the child exposed two
related replay problems: session refresh could discard the buffered
event, and a newly created child `ChatWidget` did not know the expected
MCP server set, which could leave its startup spinner running after
every server had settled.
MCP startup diagnostics should remain visible in the thread that owns
the startup without affecting other transcripts. The protocol also needs
to support a future app-scoped MCP lifecycle where startup is not owned
by any thread.
## Reported Behavior
The [originating Slack
report](https://openai.slack.com/archives/C08JZTV654K/p1780604538859939)
called out that using subagents could turn MCP startup failures into a
wall of yellow CLI warnings because repeated failures were not
deduplicated. The intended behavior is for those diagnostics to remain
visible once in the thread that owns the startup, without polluting the
parent transcript.
## What Changed
- add nullable `threadId` ownership to `mcpServer/startupStatus/updated`
- populate it from the app-server conversation ID for the current
thread-scoped lifecycle and regenerate the protocol schema and
TypeScript artifacts
- treat a missing or null `threadId` as app-scoped without injecting it
into the active chat transcript
- route and buffer thread-owned MCP startup notifications by thread in
the TUI
- preserve buffered MCP startup events across child session refresh
- seed expected MCP servers before replaying a thread snapshot so
startup reaches its terminal state
- suppress an identical repeated failure warning for the same server
within one startup round
The owning thread still renders the detailed failure and final `MCP
startup incomplete (...)` summary.
## How to Test
1. Configure an optional MCP server named `smoke` that exits during
initialization.
2. Launch the TUI with multi-agent support enabled.
3. Confirm the main thread's own startup failure renders one detailed
`smoke` warning and one incomplete-startup summary.
4. Spawn exactly one subagent.
5. Confirm the parent transcript does not receive the subagent's MCP
startup failure.
6. Switch to the subagent thread and confirm it contains exactly one
detailed `smoke` failure and one incomplete-startup summary.
7. Confirm the subagent's MCP startup spinner disappears and the thread
remains usable.
8. Switch between the parent and subagent and confirm the warnings
neither move nor duplicate.
Targeted tests:
- `just test -p codex-app-server-protocol`
- `just test -p codex-app-server
thread_start_emits_mcp_server_status_updated_notifications`
- `just test -p codex-tui mcp_startup`
The parent/child behavior and spinner completion were also exercised
manually in tmux. `just argument-comment-lint` was attempted but blocked
by an unrelated local Bazel LLVM empty-glob failure; touched Rust
callsites were inspected manually.
## Why
Interactive `codex resume` and `codex fork` expose both a session ID
positional and an initial prompt positional. With `--last`, Clap still
assigns the first positional to the session ID, so a command such as
`codex fork --last "/compact focus on auth"` either fails parsing or
attempts to look up the prompt as a session ID instead of sending it to
the latest session.
This makes it impossible to select the latest session and immediately
provide a follow-up prompt, even though `codex exec resume --last`
already supports that workflow.
<img width="1746" height="1024" alt="CleanShot 2026-06-06 at 17 00
47@2x"
src="https://github.com/user-attachments/assets/86885c07-a23c-48ee-b0ee-47f2484f6eb7"
/>
## What Changed
- Reinterpret the first positional as the initial prompt when
interactive `resume --last` or `fork --last` is used and no explicit
second prompt was parsed.
- Preserve the existing `resume SESSION_ID PROMPT` and `fork SESSION_ID
PROMPT` behavior.
- Add parser-level regression coverage for latest-session and
explicit-session prompt forms.
## How to Test
1. Start an interactive session, exit it, then run `codex resume --last
"continue from the latest session"`.
2. Confirm Codex resumes the latest session and submits the supplied
prompt instead of treating it as a session ID.
3. Run `codex fork --last "take a different approach"`.
4. Confirm Codex forks the latest session and submits the supplied
prompt.
5. Also verify `codex resume SESSION_ID "continue here"` and `codex fork
SESSION_ID "branch here"` still target the explicit session and submit
the prompt.
Targeted tests:
- `just test -p codex-cli` (267 passed)
## Why
This PR fixes approval sandbox semantics in the unified-exec path. The
zsh-fork runtime exposed the bug because the shell can do meaningful
work before any intercepted child `execv(2)` exists: redirections,
builtins, globbing, and pipeline setup all happen in the launch process.
If the model requested `sandbox_permissions=require_escalated`, or an
exec-policy `allow` rule explicitly bypassed the sandbox, that approved
sandbox decision needs to be preserved for the launch path and for
intercepted execs that use the same approval machinery.
The behavior is not only about zsh fork. The production changes are in
shared approval/escalation code, so they also affect non-zsh-fork
intercepted exec paths that go through the same sandbox decision logic.
The narrow intent is to preserve the approval decision while still
keeping denied-read profiles and bounded additional-permission requests
sandboxed.
## Production Changes
- `codex-rs/core/src/tools/runtimes/unified_exec.rs`: derives a
`launch_sandbox_permissions` value from the requested sandbox
permissions and the runtime filesystem policy, then uses that value for
managed-network/env setup and launch sandbox selection. This keeps full
approval or policy-bypass decisions visible to the first unified-exec
attempt, while still preventing a full sandbox override from discarding
denied-read restrictions. Direct unified exec keeps the same decision
surface; the important difference is that zsh-fork launch setup no
longer accidentally loses the approved parent sandbox decision.
- `codex-rs/core/src/tools/runtimes/shell/unix_escalation.rs`: makes
intercepted-exec escalation selection explicit for the three sandbox
permission modes. `UseDefault` only escalates when an exec-policy
decision allows sandbox bypass, `RequireEscalated` escalates when
unsandboxed execution is allowed, and `WithAdditionalPermissions`
escalates through the bounded additional-permissions path instead of
being treated as a full unsandboxed override. Unsandboxed intercepted
execs now also rebuild the environment as `RequireEscalated`, which
strips managed-network proxy variables consistently with other
unsandboxed execution.
## Test Coverage
Most of the PR is tests. The new coverage verifies:
- unified exec preserves parent approval and exec-policy sandbox
decisions for zsh-fork launch selection;
- bounded `with_additional_permissions` remains sandboxed and
permission-profile based;
- denied-read profiles are not weakened by parent approval;
- explicit prompt rules still prompt for intercepted execs after the
parent command is approved;
- unsandboxed intercepted execs strip managed-network env vars.
No documentation update is needed; this is an internal approval/sandbox
correctness fix.
---
[//]: # (BEGIN SAPLING FOOTER)
Stack created with [Sapling](https://sapling-scm.com). Best reviewed
with [ReviewStack](https://reviewstack.dev/openai/codex/pull/24981).
* #24982
* __->__ #24981
## Why
A remote-control WebSocket handshake can receive a generic HTTP 404 when
an intermediary routes the request without preserving the WebSocket
upgrade. Treating every 404 as proof that the remote app server is gone
clears valid enrollment and causes repeated re-enrollment, new
environment and server IDs, Habitat churn, and noisy `/server/enroll`
traffic.
## What Changed
- Clear enrollment only when a 404 JSON response explicitly contains
`{"detail":"Remote app server not found"}`.
- Preserve enrollment for empty, plain-text, malformed, or otherwise
unrecognized 404 responses, return the transport error, and retry with
the existing reconnect backoff.
- Log the status, correlation headers (`request-id` or
`x-oai-request-id`, plus `cf-ray`), and bounded/redacted response body
for unrecognized 404s.
- Cover both explicit missing-server re-enrollment and generic 404
enrollment preservation/reconnect behavior.
## Verification
`just test -p codex-app-server-transport` passes all 114 tests on the
rebased branch, including the targeted explicit and generic WebSocket
404 scenarios.
Related issue: N/A
Skill reloads can get noisy when the watcher keeps triggering
`skills/list` and the same invalid `SKILL.md` error comes back each
time.
This keeps the first warning visible, then suppresses repeats while the
same `(path, message)` is still active. If the error clears and later
comes back, or if the message changes, it will show again.
Validation:
- `just fmt`
- `just test -p codex-tui skill_load_warning_state`
## Why
Permission profile allowlists are an enterprise security boundary, but
they also need to compose across the managed requirements layers added
in #24620.
A map representation lets each requirements layer add, allow, or revoke
individual profiles without replacing an entire array.
## Managed Contract
Administrators configure the mergeable allow map with
`allowed_permission_profiles`. A recommended enterprise configuration
explicitly lists every built-in and custom profile users should be able
to select:
```toml
default_permissions = "review_only"
[allowed_permission_profiles]
":read-only" = true
":workspace" = true
review_only = true
# ":danger-full-access" is intentionally omitted, so it is denied.
[permissions.review_only]
extends = ":read-only"
```
- Profiles whose effective merged value is `true` are allowed.
- Missing profiles and profiles set to `false` are denied.
- This is a closed allowlist: built-in profiles and profiles introduced
in future versions are denied unless explicitly allowed.
- Explicitly list each built-in profile the enterprise wants to make
available. Omit built-ins such as `:danger-full-access` when they should
remain unavailable.
- Set `default_permissions` explicitly to the allowed profile users
should receive when they have no local selection.
- Higher-precedence layers override only the profile keys they define.
- `false` is only needed when a higher-precedence layer must revoke a
`true` inherited from a lower layer.
- Explicit keys must refer to known built-in or managed profiles.
A custom or narrowed allowlist requires an allowed
`default_permissions`. For compatibility, if both `:workspace` and
`:read-only` are explicitly allowed, an omitted default resolves to
`:workspace`; customer configurations should still set the intended
default explicitly.
When `allowed_permission_profiles` is absent, existing implicit
permission and legacy `sandbox_mode` behavior is unchanged.
## What Changed
- Add `allowed_permission_profiles` as a `BTreeMap<String, bool>` that
merges per profile across requirements layers.
- Enforce managed defaults, strict denial of omitted profiles, and the
explicitly allowed standard-pair fallback.
- Expose `allowedPermissionProfiles` through `configRequirements/read`
and regenerate its schemas.
- Add regression coverage for map composition and revocation, managed
defaults, strict denial of omitted built-ins, and API output.
## Verification
- Focused `codex-config` coverage for layered map composition and
revocation
- Focused `codex-core` coverage for managed defaults, invalid defaults,
strict denial of omitted built-ins, and the standard built-in pair
- Focused `codex-app-server` coverage for requirements API output
- Scoped Clippy for `codex-config`, `codex-core`,
`codex-app-server-protocol`, and `codex-app-server`
## Documentation
The managed `requirements.toml` documentation should introduce
`allowed_permission_profiles` as a closed permission-profile allowlist
before this setting is published on developers.openai.com.
---------
Co-authored-by: Codex <noreply@openai.com>
## Summary
- send `X-OpenAI-Internal-Codex-Responses-Lite: true` on HTTP Responses
requests and WebSocket upgrade requests when model metadata enables
Responses Lite
- use client metadata when sending it over the websocket
This PR is stacked on #26490.
## Why
The Responses Lite marker is request-scoped for HTTP but
connection-scoped for Responses-over-WebSocket because it is carried on
the upgrade request. Reusing a cached socket opened for the opposite
mode would therefore send the wrong transport contract.
## Validation
- `just test -p codex-core responses_lite`
- `just test -p codex-core
responses_websocket_reconnects_when_responses_lite_mode_changes`
- `just fix -p codex-core`
- `just fmt`
## Summary
- add v2 personal access token support for `codex login
--with-access-token` and `CODEX_ACCESS_TOKEN`
- classify opaque `at-` tokens separately from legacy Agent Identity
JWTs
- hydrate required ChatGPT account metadata through AuthAPI
`/v1/user-auth-credential/whoami`
- use PATs directly as bearer tokens while preserving existing ChatGPT
account surfaces
- expose PAT-backed auth as the explicit `personalAccessToken`
app-server auth mode
## Implementation
PAT auth is intentionally small and stateless. Loading a PAT performs
one AuthAPI metadata request, stores the hydrated metadata in the
in-memory auth object, and redacts the secret from debug output. Legacy
Agent Identity JWT handling remains unchanged. The shared access-token
classifier lives in a private neutral module because it dispatches
between both credential types.
PAT hydration fails closed when AuthAPI omits any required metadata,
including email. Hydrated metadata is intentionally not persisted:
startup performs a live `whoami` preflight so revoked tokens or changed
account metadata are not accepted from a stale cache.
## Workspace restriction scope
This change intentionally does **not** apply
`forced_chatgpt_workspace_id` to PAT authentication. The setting is a
client-side config guardrail, not an authorization boundary, and PAT
does not currently require workspace-ID parity. The PAT login and
`CODEX_ACCESS_TOKEN` paths therefore validate through AuthAPI without
threading workspace-restriction state through access-token loading.
Existing workspace checks for non-PAT auth remain on their established
paths.
## App-server compatibility
The public app-server `AuthMode` is shared across v1 and v2, and
PAT-backed auth reports `personalAccessToken` through both APIs.
Following human review, this intentionally removes the temporary v1
compatibility mapping that reported PATs as `chatgpt`; the deprecated v1
API is kept in parity with v2 rather than maintaining a separate closed
enum. Clients with exhaustive auth-mode handling in either API version
must add the new case and should generally treat it as ChatGPT-backed
unless they need PAT-specific behavior.
The v1 auth-status response still omits the raw PAT when `includeToken`
is requested because that response cannot carry the account metadata
needed to reuse the credential safely. Persisted PAT auth also omits the
new enum value so older Codex builds can deserialize `auth.json` and
infer PAT auth from the credential field after a rollback.
## Validation
Latest review-fix validation:
- `CARGO_INCREMENTAL=0 just test -p codex-login` (126 passed)
- `CARGO_INCREMENTAL=0 just test -p codex-cli` (263 passed)
- `CARGO_INCREMENTAL=0 just test -p codex-cli
stored_auth_validation_handles_personal_access_token`
- `CARGO_INCREMENTAL=0 just test -p codex-app-server-protocol` (226
passed)
- `CARGO_INCREMENTAL=0 just test -p codex-models-manager
refresh_available_models_uses_remote_only_catalog_for_chatgpt_auth`
- `CARGO_INCREMENTAL=0 just test -p codex-tui
existing_non_oauth_chatgpt_login_counts_as_signed_in`
- `CARGO_INCREMENTAL=0 just fix -p codex-login -p
codex-app-server-protocol -p codex-models-manager -p codex-tui -p
codex-cli`
- `just fmt`
- `git diff --check`
The broader `codex-tui` suite previously compiled and ran 2,834 tests.
Three unrelated environment-sensitive guardian/IDE-socket tests failed
after retries; the PAT-relevant TUI coverage passed.
## Summary
- add `Feature::TerminalVisualizationInstructions` as
`UnderDevelopment`, disabled by default
- keep terminal visualization instructions inside the TUI package
- append them to existing developer instructions for TUI start, resume,
and fork flows only when enabled
- intentionally do not apply them to `codex exec`
## Rollout
Control behavior is unchanged. TUI dogfooders can enable
`terminal_visualization_instructions`; no default user receives the new
terminal-specific instructions.
The shared visualization-selection rule is supplied separately through
the `codex_proxy_model_3` Statsig layer for every target Codex model
slug in the gated cohort. This TUI feature determines how to render an
appropriate visualization on the terminal surface; the model-layer
treatment determines when to use one.
## Validation
- `cargo test -p codex-tui
terminal_visualization_instructions_are_gated_for_all_tui_thread_flows
--lib`
- `cargo test -p codex-features --lib`
- `cargo fmt --all -- --check`
- `git diff --check`
- GPT-5.4 and GPT-5.5 real prompt-pipeline smoke tests: both visualized
the positive mapping case, abstained on the negative route case, and
passed exact prompt-stack verification on CLI and App
- refreshed onto current `main` with a clean merge and reran the focused
validation
The full 53-probe all-model treatment comparison and requested
production coding evals remain rollout gates before broadening beyond
the initial employee cohort.
This PR remains open for normal human review.
## Summary
Responses Lite does not execute hosted Responses tools, so models using
it must route web search and image generation through Codex-owned
executors & standalone Response's API endpoints.
This PR is stacked on #26487.
## Validation
- `cargo test -p codex-core responses_lite_ --lib`
- `cargo test -p codex-core
standalone_executors_remain_hidden_without_flags_or_responses_lite
--lib`
- `cargo test -p codex-core
hosted_tools_follow_provider_auth_model_and_config_gates --lib`
- `cargo test -p codex-web-search-extension -p
codex-image-generation-extension`
- `cargo test -p codex-app-server --test all standalone_`
- `cargo fmt --all -- --check`
## Summary
- Remove the legacy startup remote plugin sync path that called
`/plugins/list` and reconciled curated plugin cache/config.
- Remove the `sync_plugins_from_remote` API, its result/error types,
startup marker task, and tests that expected the legacy request.
- Keep the current remote installed bundle sync and remote catalog flows
(`/ps/plugins/installed` and `/ps/plugins/list`) intact.
## Validation
- `just fmt`
- `git diff --check`
- `env HOME=/private/tmp/codex-xin-build-home
USERPROFILE=/private/tmp/codex-xin-build-home just test -p
codex-core-plugins`
- Searched for legacy `/plugins/list` sync references; remaining matches
are `/ps/plugins/list` catalog tests/code.
## Notes
- `just test -p codex-app-server plugin_list` is currently blocked
before running filtered tests by an unrelated compile error in
`app-server/tests/suite/v2/image_generation.rs`:
`app_test_support::McpProcess` is not exported.
## Why
Terminal turn errors can leave a goal active. Automatic goal
continuation may then repeatedly hit a permanent failure, including
compaction requests rejected with HTTP 400, and consume excessive
tokens.
This PR changes the goal extension to treat all turn-ending errors
(including non-retryable errors and retryable errors that have exceeded
their retry count) as "blocking" for the goal. The downside to this
change is that there are some errors that may eventually succeed (e.g. a
429 due to a service outage), and previously the goal runtime would have
kept the agent going in these situations.
## What changed
- Block the current active goal when a turn ends with an error other
than a usage-limit error.
- Preserve the existing `usage_limited` transition for usage-limit
errors.
- Share progress accounting, guarded state updates, metrics, and event
emission in the goal runtime.
## Why
Submission-side `Op` payloads are now an internal handoff inside the
Rust codebase, so keeping a stable serde contract there adds complexity
without a real wire consumer.
## What changed
- remove serde/schema annotations from `Submission`, `Op`, and
submission-only payload types like thread settings overrides, additional
context, realtime conversation params, `TurnEnvironmentSelection`, and
`RequestUserInputResponse`
- delete the `Op` serialization tests and the now-unused double-option
prompt serde helper
- keep event/API-facing serialization where it is still required, and
serialize the `request_user_input` tool output from its wire payload
instead of the core response struct
- update `protocol_v1.md` to call out that events remain the serialized
transport surface while submission payloads are implementation details
## Testing
- `just test -p codex-protocol`
- `cargo check -p codex-core -p codex-app-server -p codex-thread-store`
- `just test -p codex-core request_user_input`
## Stack
1. [#26547](https://github.com/openai/codex/pull/26547) - [1 of 2] Align
goal extension with core behavior
2. [#26548](https://github.com/openai/codex/pull/26548) - [2 of 2] Move
goal runtime to extension
## Why
This PR completes the switch of the goal behavior to the
extension-backed runtime and removes the old core goal implementation.
## What Changed
- Installs the goal extension for app-server `ThreadManager` sessions.
- Routes app-server thread goal `get`, `set`, and `clear` through
`GoalService`.
- Uses thread-idle lifecycle emission after goal resume and snapshot
ordering so the extension can decide whether to continue the goal.
- Forwards extension goal updates through a FIFO async app-server
notification path so backpressure does not drop them or reorder updates.
- Keeps review turns from enabling goal runtime behavior.
- Plans extension tools before dynamic tools so built-in goal tool names
keep their old precedence when goals are enabled.
- Removes the old core goal runtime, core goal tool handlers, and core
goal tool specs.
- Updates tests that were coupled to the core-owned goal runtime while
leaving the legacy `<goal_context>` compatibility path in core for old
threads.
- Removes the stale cargo-shear ignore now that `codex-goal-extension`
is used by the workspace.
- Keeps realtime event matching exhaustive after removing the old
goal-specific realtime text path.
## Validation
- Ran manual `/goal` runs in TUI. Validated time accounting matched
wall-clock time and goal lifecycle state transitions.
## Context
The installed-app suggestion expansion added in #24996 reads plugin
details for trusted file-backed marketplace candidates because the list
response does not include app ids. On Windows-backed WSL mounts, the
local `openai-curated` checkout lives under `$CODEX_HOME/.tmp/plugins`,
and those per-plugin detail reads can be very slow.
Remote curated already has cached app ids, so it does not need the same
local filesystem traversal.
## Summary
- Keep only the WSL Windows-backed local `openai-curated` checkout on
the legacy fallback/configured discovery path.
- Preserve installed-app expansion for non-WSL file-backed marketplaces
and remote curated.
- Add focused tests for the WSL local curated path predicate.
## Test
- `just test -p codex-core-plugins discoverable`
- `just test -p codex-core plugins::discoverable::tests`
## Summary
TUI startup loads related plugin data from `hooks/list`, session MCP
initialization, and plugin skill warmup. These paths repeated filesystem
discovery and emitted the same plugin warnings, while `hooks/list` and
account/model bootstrap ran serially.
This change:
- Reuses one immutable plugin load outcome across startup consumers.
- Keys the cache only on plugin-relevant configuration.
- Single-flights concurrent plugin loads and prevents invalidated loads
from repopulating the cache.
- Runs hook discovery and account/model bootstrap concurrently.
- Preserves configuration-migration ordering, hook review behavior, and
accurate startup telemetry.
In 10 alternating release-build launches in the Ruff repository with the
existing `~/.codex` configuration, median time to the first editable
composer decreased from 833ms to 504ms. The branch was faster in 9 of 10
pairs, with a paired median improvement of 312ms.
## Summary
`codex resume --last` currently lists sessions by updated time using
scan-and-repair. Updated-time filesystem listing must stat every rollout
before applying the cwd, provider, and source filters, so startup scales
with the entire local session history...
This change queries the state DB first for the latest matching session.
For local workspaces, we only accept the indexed result when its rollout
path still exists; otherwise we retry with scan-and-repair. The same
lookup path is shared by `fork --last`.
I benchmarked the same `thread/list` request used by `resume --last` in
my local `ruff` checkout against a Codex home with 2,599 active rollouts
totaling 3.7 GiB, including 90 Ruff threads.
Across five fresh release app-server processes with warm filesystem
caches, the state-DB-only lookup had median latency of 0.37-0.44 ms,
while scan-and-repair had median latency of 139-162 ms. First-request
latency was 0.7-1.7 ms versus 142-185 ms.
So this **removes roughly 140-160 ms from the `resume --last` lookup**
on this machine, and makes that lookup over 300x faster.
The tradeoff is that this does leave two correctness gaps:
- If a newer matching rollout is missing from SQLite but an older
matching row exists, the fast path resumes the older thread and never
falls back to the filesystem scan.
- If an existing row has stale filter or ordering metadata, the fast
path can select a different thread from scan-and-repair. The rollout
tests already demonstrate this for stale cwd metadata: state-DB-only
returns the stale match, while scan-and-repair removes and repairs it.
So you could end up seeing the "wrong" result in cases like...
1. A crash or SQLite error occurs between Codex writing the conversation
file and updating SQLite, leaving the newer file unindexed.
2. An older Codex version, restore, or manual copy adds a conversation
file after SQLite’s one-time backfill completed.
These seem pretty rare though (and sessions can always be recovered via
other mechanisms -- `--last` is just a convenience feature), and I think
the tradeoffs are good here?
Stacked on #26532.
## Why
#26532 moves cwd normalization to the app-server/core boundary.
`runtimeWorkspaceRoots` still accepted raw paths in v2 requests and in
`ConfigOverrides`, which left core responsible for interpreting those
roots later. This makes runtime workspace roots follow the same
absolute-path boundary as cwd.
## What
- Change v2 `runtimeWorkspaceRoots` request fields for `thread/start`,
`thread/resume`, `thread/fork`, and `turn/start` to `AbsolutePathBuf`.
- Deduplicate already-absolute runtime roots in app-server handlers and
pass them through `ConfigOverrides.workspace_roots` as
`AbsolutePathBuf`.
- Update TUI and exec client request builders to pass absolute runtime
roots directly.
- Update app-server docs, schema fixtures, and focused tests for
absolute runtime roots.
## Testing
- `just test -p codex-app-server-protocol`
- `just test -p codex-app-server runtime_workspace_roots`
- `just test -p codex-core
session_permission_profile_rebinds_runtime_workspace_roots`
- `just test -p codex-tui app_server_session`
- `just test -p codex-exec`
## Summary
Add flat profiling fields to `codex_turn_event` so analytics can explain
where turn wall-clock time is spent without changing tool execution
behavior.
The profile reports:
- time before the first sampling request
- sampling time across all attempts and follow-ups
- overhead between sampling requests
- time blocked in the post-sampling tool drain
- time after the final sampling request
- sampling request and retry counts
## Implementation
- Extend the existing turn timing state with constant-memory phase
accounting and one RAII phase guard.
- Observe sampling and the existing post-sampling drain only at turn
orchestration boundaries.
- Keep tool runtime, tool futures, response item handling, and turn
lifecycle values unchanged.
- Add the profiling fields directly to the existing analytics turn event
without changing app-server protocol or rollout persistence.
- Use the existing turn `status` to distinguish completed, failed, and
interrupted profiles.
Exact sampling/tool overlap is intentionally omitted because measuring
tool completion accurately would require hooks in the tool execution
path.
## Validation
- Add app-server end-to-end coverage for a single-sampling turn with no
blocking tool work.
- Add app-server end-to-end coverage for `request_user_input` blocking
followed by a second sampling request.
- CI is running on the PR; tests were not executed locally per
repository guidance.
## Why
Windows managed filesystem permissions can now be backed by a real
Windows sandbox. `exec-policy` was still treating the managed read-only
policy shape as if there were never a sandbox backend, so benign
unmatched commands such as PowerShell directory listings could be
rejected with `blocked by policy` even when `windows.sandbox` was
enabled.
The inverse case still needs to stay conservative: when the Windows
sandbox backend is disabled, managed filesystem restrictions are only
configuration intent, not an enforced filesystem boundary. That applies
to writable-root restricted profiles too, not just read-only profiles.
## What Changed
- Thread the effective `WindowsSandboxLevel` into exec-policy approval
decisions for shell, unified exec, and intercepted shell exec paths.
- Treat managed restricted filesystem profiles as lacking sandbox
protection only on Windows when `WindowsSandboxLevel::Disabled`.
- Exclude full-disk-write profiles from that no-backend path because
they do not rely on filesystem sandbox enforcement.
- Remove the cwd-sensitive read-only heuristic and the now-stale cwd
plumbing from exec-policy approval contexts.
- Add Windows coverage for both enabled-sandbox and disabled-backend
behavior, including a writable-root managed profile.
## Validation
- Added/updated `exec_policy` coverage for managed filesystem
restrictions, full-disk-write exclusion, enabled Windows sandbox
behavior, and disabled-backend read-only/writable-root behavior.
- `just test -p codex-core exec_policy` — 100 passed, 10 leaky
- Empirical local `codex exec` probe with `--sandbox read-only -c
'windows.sandbox="unelevated"'`: PowerShell directory listing completed
successfully.
- Disabled-backend control with Windows sandbox cleared: the same
command was rejected with `blocked by policy`.
## Why
Pressing `Esc` on a turn that produced no visible output restores the
submitted prompt so the user can keep editing it. That restore path
preserved the prompt content, images, and mention bindings, but left the
composer cursor at the start of the restored text. The next edit
therefore inserted at the beginning instead of continuing from the end
of the prompt.
## What Changed
- Move the cursor to the end after
`BottomPane::set_composer_text_with_mention_bindings` rehydrates a
restored draft.
- Add test-only cursor accessors so restore tests can assert the
composer state directly.
- Extend the queued restore regression to assert the restored composer
cursor is positioned at `text.len()`.
## How to Test
Manual reviewer flow:
1. Start Codex in the TUI.
2. Submit a prompt that will take long enough to interrupt.
3. Press `Esc` before any visible assistant output appears.
4. Confirm the prompt is restored into the composer and the cursor is at
the end, so typing appends to the prompt.
5. Repeat with a prompt that includes an attached image or resolved
mention and confirm the restored content remains intact.
Targeted tests:
- `just test -p codex-tui
chatwidget::tests::composer_submission::queued_restore_with_remote_images_keeps_local_placeholder_mapping`
Lint note:
- `just argument-comment-lint` is blocked locally by the existing Bazel
`compiler-rt` empty glob failure before analyzing touched code. The
touched Rust diff was manually inspected and adds no new opaque
positional literal callsites.
## Why
On Windows, the TUI could not shade the composer against the terminal
background because `terminal_palette::default_colors()` always fell back
to `None`. That preserved safety, but it also meant terminals that do
support OSC 10/11 default color replies had no path to report their real
background color.
This keeps the existing fallback behavior for unsupported terminals
while allowing capable Windows terminals to report their default
foreground/background colors during startup.
| Before | After |
|---|---|
| <img width="1235" height="658" alt="win-before"
src="https://github.com/user-attachments/assets/ff756589-fcb3-43de-8f2a-ebc0369b30dd"
/> | <img width="1235" height="658" alt="win-after"
src="https://github.com/user-attachments/assets/9563ff20-4be5-4608-9414-a2afb647e745"
/> |
## What Changed
- Moved the OSC 10/11 default color parser in
`tui/src/terminal_probe.rs` out of the Unix-only implementation so it
can be reused by Windows.
- Added a Windows-only bounded OSC 10/11 probe using raw console handles
and the existing `windows-sys` dependency.
- Added Windows palette caching in `tui/src/terminal_palette.rs` so
startup probe results, including `None`, are reused instead of probing
again later.
- Wired the Windows color probe into TUI startup after the existing
non-Unix crossterm cursor and keyboard checks.
- Added parser coverage for malformed, partial, and noisy OSC color
replies.
If the probe fails, times out, receives only one color, or receives
malformed data, the cache stores `None` and the composer keeps the
current behavior.
## How to Test
1. On Windows, start Codex in a terminal that supports OSC 10/11 default
color replies.
2. Open the TUI composer.
3. Confirm the composer/status area is painted using the terminal's
reported default background, instead of leaving the background unshaded.
4. Start Codex in a terminal that does not answer OSC 10/11, or
otherwise blocks terminal color replies.
5. Confirm startup still succeeds and the composer uses the existing
fallback behavior.
Targeted tests:
- `CARGO_TARGET_DIR=/private/tmp/codex-windows-osc-default-colors-target
just test -p codex-tui terminal_probe`
Additional local verification:
- `CARGO_TARGET_DIR=/private/tmp/codex-windows-osc-default-colors-target
just test -p codex-tui` was run; 2774 tests passed, and two unrelated
Guardian feature-flag tests failed reproducibly when isolated.
- `just argument-comment-lint` was attempted but blocked by the local
Bazel/LLVM `include/sanitizer/*.h` empty glob issue. Touched Rust
literal callsites were inspected manually.
- `cargo check -p codex-tui --target x86_64-pc-windows-msvc` was
attempted after installing the target, but local macOS cross-checking is
blocked by missing Windows C SDK headers in native dependencies
(`ring`/`aws-lc-sys`).
---------
Co-authored-by: Kevin Bond <kbond@openai.com>
## Stack
1. [#26547](https://github.com/openai/codex/pull/26547) - [1 of 2] Align
goal extension with core behavior
2. [#26548](https://github.com/openai/codex/pull/26548) - [2 of 2] Move
goal runtime to extension
## Why
The goal runtime is moving out of `codex-core` and into
`codex-goal-extension`. This first PR brings the extension back in line
with the current core behavior before the follow-up PR switches
app-server sessions over to the extension, so that review can focus on
ownership and wiring rather than hidden behavior drift.
## What Changed
- Updates the extension `create_goal` and `update_goal` tool
schemas/descriptions to match the current core wording for explicit
token budgets, blocked-goal audits, resumed blocked goals, and
system-owned budget/usage-limit transitions.
- Marks `codex-goal-extension` as the live `/goal` extension crate
rather than an unwired sketch.
- Looks up the live thread before reading goal state for idle
continuation, so continuation setup exits early when no live thread can
accept the automatic turn.
## What
Exposes the pairing status transport as experimental app-server v2 RPC
`remoteControl/pairing/status`.
- Adds request/response protocol types for exactly one lookup key:
`pairingCode` or `manualPairingCode`, returning `{ claimed }`.
- Registers the RPC with `global_shared_read("remote-control-pairing")`.
- Wires the method through `MessageProcessor` and
`RemoteControlRequestProcessor`.
- Validates missing/conflicting pairing-code params as invalid requests.
- Documents the RPC in `app-server/README.md`.
- Adds processor, protocol export, and JSON-RPC integration coverage for
both code paths.
## Why
This is the app-server surface the desktop app can poll while the
QR/manual pairing modal is active.
Depends on https://github.com/openai/codex/pull/26449
Related backend change: https://github.com/openai/openai/pull/990244
## Verification
- `cargo test --manifest-path app-server-protocol/Cargo.toml
remote_control`
- `cargo test --manifest-path app-server/Cargo.toml remote_control`
- `cargo fmt --all --check`
- `git diff --check`
## Summary
During assistant-message streaming, blank markdown lines in the
transient active tail were prefixed with two spaces. Ratatui measured
those whitespace-only lines as two viewport rows, so list- and
table-heavy answers showed doubled vertical gaps while streaming and
then visibly compacted when finalized into scrollback.
- keep whitespace-only `StreamingAgentTailCell` lines structurally empty
while preserving nonblank message prefixes
- clear impossible hyperlink metadata when normalizing a blank tail line
- add an inline snapshot and height regression proving one blank
markdown line occupies one viewport row
Related to #26618, but fixes a separate live-tail row-height issue
rather than stale committed markdown content.
## How to Test
Recommended before/after reproduction:
1. Start the latest Codex build without this change.
2. Submit this exact prompt:
> Send 20 different lists: bullets vs numbered, simple vs complex with
paragraphs in between items, etc. Intertwine them with some tables and
some paragraphs.
3. While the answer streams, observe duplicated vertical gaps around
list items and paragraphs. When the answer finishes, observe the spacing
compact.
4. Start this branch with `just c` and submit the same prompt.
5. Confirm each intended blank markdown line occupies one terminal row
throughout streaming and that the spacing does not compact or jump when
the answer finishes.
6. As a focused regression, verify the sections after the first table,
especially loose lists with paragraphs between items; those blank rows
should remain stable throughout streaming.
Targeted tests:
- `just test -p codex-tui
streaming_agent_tail_blank_line_uses_one_viewport_row`
- `just test -p codex-tui history_cell::tests`
## Test Notes
- Verified the exact prompt above in a real tmux TUI using latest Codex
and this branch as the before/after comparison.
- The full `just test -p codex-tui` run completed 2,782 of 2,784 tests
successfully. Two unrelated guardian feature-flag tests fail
reproducibly in isolation because the expected `OverrideTurnContext`
message is absent.
- `just argument-comment-lint` is blocked locally by the existing Bazel
`compiler-rt` missing-header glob error; the touched Rust diff was
inspected manually for opaque positional literals.
## Why
Turn diffs were tracked as one flat set of absolute paths. In
multi-environment turns, local and remote environments can report the
same path while representing different filesystems, so a single path key
can collapse distinct changes or attribute them to the wrong
environment.
The environment name is **NOT** included in the generated unified diff.
This can come later.
## What
Adds transport support for checking remote-control pairing status
against the backend.
- Adds the normalized `server/pair/status` backend URL.
- Adds backend request/response structs for exactly one lookup key:
`pairing_code` or `manual_pairing_code`, returning `{ claimed }`.
- Adds `RemoteControlEnrollment::pairing_status` and
`RemoteControlHandle::pairing_status`.
- Preserves auth refresh/retry behavior and backend error mapping.
- Adds transport coverage for pending, claimed, manual-code payloads,
token refresh, mapped backend errors, malformed responses, and URL
normalization.
## Why
Desktop needs a host-authenticated way to poll whether a QR or manual
pairing code has been claimed.
Related backend change: https://github.com/openai/openai/pull/990244
## Verification
- `cargo test --manifest-path app-server-transport/Cargo.toml
remote_control::tests::pairing_tests`
- `cargo fmt --all --check`
- `git diff --check`
## Why
Some Linux environments expose `bash` at `/usr/bin/bash` instead of
`/bin/bash`. The shell detection fallback list should cover both
standard locations once PATH/user-shell probing fails.
Stacked on #26480.
## What changed
- Add `/usr/bin/bash` to the bash fallback path list in
`codex-shell-command`.
- Extend shell type detection coverage for `/usr/bin/bash`.
- Add AGENTS.md testing guidance to avoid tests for statically defined
values and negative tests for removed logic.
## Verification
- `just test -p codex-shell-command`
## Summary
- allow `socketpair(AF_UNIX, ...)` in the proxy-routed Linux seccomp
mode
- continue denying `socket(AF_UNIX, ...)` so user commands cannot create
pathname or abstract Unix sockets
- extend the managed-proxy integration test to verify both behaviors
## Root cause
`NetworkSeccompMode::ProxyRouted` treated anonymous Unix socket pairs
like externally addressable Unix sockets and returned `EPERM`. This
breaks tools that use socket pairs for local child-process IPC even
though a socket pair cannot connect outside the sandbox or bypass the
routed proxy.
`dangerously_allow_all_unix_sockets` controls Unix-socket requests
forwarded by the managed network proxy; it does not currently configure
the Linux seccomp filter. Socket pairs should not require that dangerous
setting because they are unnamed, process-local IPC.
Related but independent: #26553 fixes host proxy bridge socket path
length handling.
---------
Co-authored-by: Codex <noreply@openai.com>
## Why
Thread settings cwd overrides are expected to be resolved before they
enter core. Keeping this boundary as a plain `PathBuf` made it easy for
core/session code to keep fallback normalization and relative-path
resolution logic in places that should only receive an already-resolved
cwd.
This is intentionally the absolute-cwd-only slice: it does not change
environment selection stickiness or cwd-to-default-environment fallback
behavior.
## What changed
- Changes `ThreadSettingsOverrides.cwd`,
`CodexThreadSettingsOverrides.cwd`, and `SessionSettingsUpdate.cwd` to
use `AbsolutePathBuf`.
- Removes core-side cwd normalization/resolution from session settings
updates.
- Updates affected core/app-server test helpers and callsites to pass
existing absolute cwd values or use `abs()` helpers.
## Validation
Opening as draft so CI can start while local validation continues.
## Summary
This is the first small step toward making multi-agent v2 agents durable
logical agents whose `ThreadManager` residency is only an implementation
detail.
This PR adds a narrow v2 reload-on-delivery hook:
- If a known v2 agent target is already loaded, delivery is unchanged.
- If the target is still registered but missing from `ThreadManager`,
delivery reloads that exact v2 thread from durable rollout history
before submitting the message.
- If the target is unknown, closed, missing from storage, or not a v2
thread, delivery still fails as not found.
The reload is wired only into existing-agent delivery paths: v2
`send_message` / `followup_task`, and legacy `send_input` when its
target is a known v2 agent.
## Stack
1. **Reload on delivery**: load known unloaded v2 agents before
`followup_task`, `send_message`, or `send_input` delivery. This PR.
2. **Residency LRU**: unload idle resident v2 agents from
`ThreadManager` without making them closed or unreachable.
3. **Execution concurrency**: count active non-root turns, not logical
agents or resident idle threads.
4. **Close semantics**: make v2 close interrupt-only and leave durable
agent identity intact.
5. **Resume cleanup**: remove user-facing v2 resume semantics;
addressing an unloaded durable agent reloads it implicitly.
## Validation
- Ran `just fmt`.
- Left broader tests and clippy to CI.
## Summary
Resumed Codex App or VS Code review sessions can contain
`::code-comment` directives that the TUI previously displayed verbatim
because only rich clients interpret them.
This change rewrites valid line-start directives into readable Markdown
during assistant-message parsing, using the session working directory
for relative file paths. The fallback is applied consistently to live
messages, replayed transcripts, and resume previews while preserving
malformed directives and existing `::git-*` parsing.
## Before
The TUI exposed the raw client directive:
```text
::code-comment{title="Fix body= parsing" body="Keep role=\"tab\", ::git-stage{cwd=/tmp}, file=, and \n literal." file="/repo/src/app.ts" start=10 end=12 priority="P2"}
```
## After
The same directive is rendered as readable review feedback:
```text
- [P2] Fix body= parsing — src/app.ts:10-12
Keep role="tab", ::git-stage{cwd=/tmp}, file=, and \n literal.
```
Fixes#25658
## Why
The TUI's `/goal` usage text only advertised the objective form even
though `/goal clear`, `/goal edit`, `/goal pause`, and `/goal resume`
are implemented. This made the lifecycle controls difficult to discover
and allowed the duplicated help text to drift from actual behavior.
Fixes#25530.
## What changed
- Show the complete `/goal [<objective>|clear|edit|pause|resume]` syntax
in usage messages.
- Share one usage string across slash-command dispatch and goal-related
app messages.
- Add inline snapshot coverage for the control-command usage path.
## Summary
Fixes#26423.
On Windows, `codex app PATH` detected Codex Desktop and launched the app
shell target, then only printed a manual instruction to open the
workspace. The Desktop app already supports
`codex://threads/new?path=...`, so the CLI can open the requested
workspace directly.
This updates the Windows launcher to normalize the workspace path,
encode it into a `codex://threads/new` deep link, and open that URL when
Codex Desktop is installed. The installer fallback still opens the
Windows installer and prints the workspace path for after installation.
## Summary
TUI config writes currently wrap app-server failures with local context
like `config/batchWrite failed in TUI`, but several user-visible paths
only render the outer error. That hides the actionable app-server
message, such as validation constraints or read-only `CODEX_HOME`
failures, leaving users with a dead-end diagnostic.
This change adds a small formatter next to the TUI config write helpers
that renders the error source chain, then uses it for model persistence,
feature persistence, project trust, status line writes, hook trust, and
hook enablement.
Fixes#26077
rhan-oaiandGitHub
ningyi-oaiandGitHub
Anton PanasenkoandGitHub
Jeremy RoseandGitHub
rka-oaiandGitHub
jifandGitHub
Felipe CouryandGitHub
Michael BolinandGitHub
Channing CongerandGitHub
Eric NingandGitHub
xl-openaiandGitHub
cooper-oaiandGitHub
vie-oaiandGitHub
Eric TrautandGitHub
pakrym-oaiandGitHub
mpc-oaiandGitHub
Charlie MarshandGitHub
Ahmed IbrahimandGitHub
iceweasel-oaiandGitHub
hefuc-oaiandGitHub