An experimental flow for env var skill dependencies. Skills can now
declare required env vars in SKILL.md; if missing, the CLI prompts the
user to get the value, and Core will store it in memory (eventually to a
local persistent store)
<img width="790" height="169" alt="image"
src="https://github.com/user-attachments/assets/cd928918-9403-43cb-a7e7-b8d59bcccd9a"
/>
web_search can now be updated per-turn, for things like changes to
sandbox policy.
`SandboxPolicy::DangerFullAccess` now sets web_search to `live`, and the
default is still `cached`.
Added integration tests.
## Context
Previous work in https://github.com/openai/codex/pull/9560 only rejected
`request_user_input` in Execute and Custom modes. Since then, additional
modes
(e.g., Code) were added, so the guard should be mode-agnostic.
## What changed
- Switch the handler to an allowlist: only Plan and PairProgramming are
allowed
- Return the same error for any other mode (including Code)
- Add a Code-mode rejection test alongside the existing Execute/Custom
tests
## Why
This prevents `request_user_input` from being used in modes where it is
not
intended, even as new modes are introduced.
Reproduce with a prompt like this with collab enabled:
```
Examine the code at <some subdirectory with a deeply nested project>. Find the most urgent issue to resolve and describe it to me.
```
Existing behavior causes the top-level agent to busy wait on subagents.
### Summary
Add `isOther` to question object from request_user_input tool input and
remove `other` option from the tool prompt to better handle tool input.
## Summary
Add dynamic tool injection to thread startup in API v2, wire dynamic
tool calls through the app server to clients, and plumb responses back
into the model tool pipeline.
### Flow (high level)
- Thread start injects `dynamic_tools` into the model tool list for that
thread (validation is done here).
- When the model emits a tool call for one of those names, core raises a
`DynamicToolCallRequest` event.
- The app server forwards it to the client as `item/tool/call`, waits
for the client’s response, then submits a `DynamicToolResponse` back to
core.
- Core turns that into a `function_call_output` in the next model
request so the model can continue.
### What changed
- Added dynamic tool specs to v2 thread start params and protocol types;
introduced `item/tool/call` (request/response) for dynamic tool
execution.
- Core now registers dynamic tool specs at request time and routes those
calls via a new dynamic tool handler.
- App server validates tool names/schemas, forwards dynamic tool call
requests to clients, and publishes tool outputs back into the session.
- Integration tests
## Summary
- Keep `request_user_input` in the tool list but reject it at runtime in
Execute/Custom modes with a clear model-facing error.
- Add a session accessor for current collaboration mode and enforce the
gate in the request_user_input handler.
- Update core/app-server tests to use Plan mode for success and add
Execute/Custom rejection coverage.
## Summary
This PR consolidates base_instructions onto SessionMeta /
SessionConfiguration, so we ensure `base_instructions` is set once per
session and should be (mostly) immutable, unless:
- overridden by config on resume / fork
- sub-agent tasks, like review or collab
In a future PR, we should convert all references to `base_instructions`
to consistently used the typed struct, so it's less likely that we put
other strings there. See #9423. However, this PR is already quite
complex, so I'm deferring that to a follow-up.
## Testing
- [x] Added a resume test to assert that instructions are preserved. In
particular, `resume_switches_models_preserves_base_instructions` fails
against main.
Existing test coverage thats assert base instructions are preserved
across multiple requests in a session:
- Manual compact keeps baseline instructions:
core/tests/suite/compact.rs:199
- Auto-compact keeps baseline instructions:
core/tests/suite/compact.rs:1142
- Prompt caching reuses the same instructions across two requests:
core/tests/suite/prompt_caching.rs:150 and
core/tests/suite/prompt_caching.rs:157
- Prompt caching with explicit expected string across two requests:
core/tests/suite/prompt_caching.rs:213 and
core/tests/suite/prompt_caching.rs:222
- Resume with model switch keeps original instructions:
core/tests/suite/resume.rs:136
- Compact/resume/fork uses request 0 instructions for later expected
payloads: core/tests/suite/compact_resume_fork.rs:215
moving `web_search` rollout serverside, so need a way to explicitly
disable search + signal eligibility from the client.
- Add `x‑oai‑web‑search‑eligible` header that signifies whether the
request can have web search.
- Only attach the `web_search` tool when the resolved `WebSearchMode` is
`Live` or `Cached`.
### What
Add `WebSearchMode` enum (disabled, cached live, defaults to cached) to
config + V2 protocol. This enum takes precedence over legacy flags:
`web_search_cached`, `web_search_request`, and `tools.web_search`.
Keep `--search` as live.
### Tests
Added tests
Emit the following events around the collab tools. On the `app-server`
this will be under `item/started` and `item/completed`
```
#[derive(Debug, Clone, Deserialize, Serialize, PartialEq, JsonSchema, TS)]
pub struct CollabAgentSpawnBeginEvent {
/// Identifier for the collab tool call.
pub call_id: String,
/// Thread ID of the sender.
pub sender_thread_id: ThreadId,
/// Initial prompt sent to the agent. Can be empty to prevent CoT leaking at the
/// beginning.
pub prompt: String,
}
#[derive(Debug, Clone, Deserialize, Serialize, PartialEq, JsonSchema, TS)]
pub struct CollabAgentSpawnEndEvent {
/// Identifier for the collab tool call.
pub call_id: String,
/// Thread ID of the sender.
pub sender_thread_id: ThreadId,
/// Thread ID of the newly spawned agent, if it was created.
pub new_thread_id: Option<ThreadId>,
/// Initial prompt sent to the agent. Can be empty to prevent CoT leaking at the
/// beginning.
pub prompt: String,
/// Last known status of the new agent reported to the sender agent.
pub status: AgentStatus,
}
#[derive(Debug, Clone, Deserialize, Serialize, PartialEq, JsonSchema, TS)]
pub struct CollabAgentInteractionBeginEvent {
/// Identifier for the collab tool call.
pub call_id: String,
/// Thread ID of the sender.
pub sender_thread_id: ThreadId,
/// Thread ID of the receiver.
pub receiver_thread_id: ThreadId,
/// Prompt sent from the sender to the receiver. Can be empty to prevent CoT
/// leaking at the beginning.
pub prompt: String,
}
#[derive(Debug, Clone, Deserialize, Serialize, PartialEq, JsonSchema, TS)]
pub struct CollabAgentInteractionEndEvent {
/// Identifier for the collab tool call.
pub call_id: String,
/// Thread ID of the sender.
pub sender_thread_id: ThreadId,
/// Thread ID of the receiver.
pub receiver_thread_id: ThreadId,
/// Prompt sent from the sender to the receiver. Can be empty to prevent CoT
/// leaking at the beginning.
pub prompt: String,
/// Last known status of the receiver agent reported to the sender agent.
pub status: AgentStatus,
}
#[derive(Debug, Clone, Deserialize, Serialize, PartialEq, JsonSchema, TS)]
pub struct CollabWaitingBeginEvent {
/// Thread ID of the sender.
pub sender_thread_id: ThreadId,
/// Thread ID of the receiver.
pub receiver_thread_id: ThreadId,
/// ID of the waiting call.
pub call_id: String,
}
#[derive(Debug, Clone, Deserialize, Serialize, PartialEq, JsonSchema, TS)]
pub struct CollabWaitingEndEvent {
/// Thread ID of the sender.
pub sender_thread_id: ThreadId,
/// Thread ID of the receiver.
pub receiver_thread_id: ThreadId,
/// ID of the waiting call.
pub call_id: String,
/// Last known status of the receiver agent reported to the sender agent.
pub status: AgentStatus,
}
#[derive(Debug, Clone, Deserialize, Serialize, PartialEq, JsonSchema, TS)]
pub struct CollabCloseBeginEvent {
/// Identifier for the collab tool call.
pub call_id: String,
/// Thread ID of the sender.
pub sender_thread_id: ThreadId,
/// Thread ID of the receiver.
pub receiver_thread_id: ThreadId,
}
#[derive(Debug, Clone, Deserialize, Serialize, PartialEq, JsonSchema, TS)]
pub struct CollabCloseEndEvent {
/// Identifier for the collab tool call.
pub call_id: String,
/// Thread ID of the sender.
pub sender_thread_id: ThreadId,
/// Thread ID of the receiver.
pub receiver_thread_id: ThreadId,
/// Last known status of the receiver agent reported to the sender agent before
/// the close.
pub status: AgentStatus,
}
```
This PR is in the scope of multi-agent work.
An agent (=thread) can now spawn other agents. Those other agents are
not attached to any clients. We need a way to make sure that the clients
are aware of the new threads to look at (for approval for example). This
PR adds a channel to the `ThreadManager` that pushes the ID of those
newly created agents such that the client (here the app-server) can also
subscribe to those ones.
The description of the `shell` arg for `exec_command` states the default
is `/bin/bash`, but AFAICT it's the user's default shell.
Default logic
[here](2a06d64bc9/codex-rs/core/src/tools/handlers/unified_exec.rs (L123)).
EDIT: #9004 has an alternative where we inform the model of the default
shell itself.
Add implementation for the `wait` tool.
For this we consider all status different from `PendingInit` and
`Running` as terminal. The `wait` tool call will return either after a
given timeout or when the tool reaches a non-terminal status.
A few points to note:
* The usage of a channel is preferred to prevent some races (just
looping on `get_status()` could "miss" a terminal status)
* The order of operations is very important, we need to first subscribe
and then check the last known status to prevent race conditions
* If the channel gets dropped, we return an error on purpose
Agent wouldn't "see" attached images and would instead try to use the
view_file tool:
<img width="1516" height="504" alt="image"
src="https://github.com/user-attachments/assets/68a705bb-f962-4fc1-9087-e932a6859b12"
/>
In this PR, we wrap image content items in XML tags with the name of
each image (now just a numbered name like `[Image #1]`), so that the
model can understand inline image references (based on name). We also
put the image content items above the user message which the model seems
to prefer (maybe it's more used to definitions being before references).
We also tweak the view_file tool description which seemed to help a bit
Results on a simple eval set of images:
Before
<img width="980" height="310" alt="image"
src="https://github.com/user-attachments/assets/ba838651-2565-4684-a12e-81a36641bf86"
/>
After
<img width="918" height="322" alt="image"
src="https://github.com/user-attachments/assets/10a81951-7ee6-415e-a27e-e7a3fd0aee6f"
/>
```json
[
{
"id": "single_describe",
"prompt": "Describe the attached image in one sentence.",
"images": ["image_a.png"]
},
{
"id": "single_color",
"prompt": "What is the dominant color in the image? Answer with a single color word.",
"images": ["image_b.png"]
},
{
"id": "orientation_check",
"prompt": "Is the image portrait or landscape? Answer in one sentence.",
"images": ["image_c.png"]
},
{
"id": "detail_request",
"prompt": "Look closely at the image and call out any small details you notice.",
"images": ["image_d.png"]
},
{
"id": "two_images_compare",
"prompt": "I attached two images. Are they the same or different? Briefly explain.",
"images": ["image_a.png", "image_b.png"]
},
{
"id": "two_images_captions",
"prompt": "Provide a short caption for each image (Image 1, Image 2).",
"images": ["image_c.png", "image_d.png"]
},
{
"id": "multi_image_rank",
"prompt": "Rank the attached images from most colorful to least colorful.",
"images": ["image_a.png", "image_b.png", "image_c.png"]
},
{
"id": "multi_image_choice",
"prompt": "Which image looks more vibrant? Answer with 'Image 1' or 'Image 2'.",
"images": ["image_b.png", "image_d.png"]
}
]
```
Handle null tool arguments in the MCP resource handler so optional
resource tools accept null without failing, preserving normal JSON
parsing for non-null payloads and improving robustness when models emit
null; this avoids spurious argument parse errors for list/read MCP
resource calls.
Sort list_dir entries before applying offset/limit so pagination matches
the displayed order, update pagination/truncation expectations, and add
coverage for sorted pagination. This ensures stable, predictable
directory pages when list_dir is enabled.
Add metrics capabilities to Codex. The `README.md` is up to date.
This will not be merged with the metrics before this PR of course:
https://github.com/openai/codex/pull/8350
**Summary**
This PR makes “ApprovalDecision::AcceptForSession / don’t ask again this
session” actually work for `apply_patch` approvals by caching approvals
based on absolute file paths in codex-core, properly wiring it through
app-server v2, and exposing the choice in both TUI and TUI2.
- This brings `apply_patch` calls to be at feature-parity with general
shell commands, which also have a "Yes, and don't ask again" option.
- This also fixes VSCE's "Allow this session" button to actually work.
While we're at it, also split the app-server v2 protocol's
`ApprovalDecision` enum so execpolicy amendments are only available for
command execution approvals.
**Key changes**
- Core: per-session patch approval allowlist keyed by absolute file
paths
- Handles multi-file patches and renames/moves by recording both source
and destination paths for `Update { move_path: Some(...) }`.
- Extend the `Approvable` trait and `ApplyPatchRuntime` to work with
multiple keys, because an `apply_patch` tool call can modify multiple
files. For a request to be auto-approved, we will need to check that all
file paths have been approved previously.
- App-server v2: honor AcceptForSession for file changes
- File-change approval responses now map AcceptForSession to
ReviewDecision::ApprovedForSession (no longer downgraded to plain
Approved).
- Replace `ApprovalDecision` with two enums:
`CommandExecutionApprovalDecision` and `FileChangeApprovalDecision`
- TUI / TUI2: expose “don’t ask again for these files this session”
- Patch approval overlays now include a third option (“Yes, and don’t
ask again for these files this session (s)”).
- Snapshot updates for the approval modal.
**Tests added/updated**
- Core:
- Integration test that proves ApprovedForSession on a patch skips the
next patch prompt for the same file
- App-server:
- v2 integration test verifying
FileChangeApprovalDecision::AcceptForSession works properly
**User-visible behavior**
- When the user approves a patch “for session”, future patches touching
only those previously approved file(s) will no longer prompt gain during
that session (both via app-server v2 and TUI/TUI2).
**Manual testing**
Tested both TUI and TUI2 - see screenshots below.
TUI:
<img width="1082" height="355" alt="image"
src="https://github.com/user-attachments/assets/adcf45ad-d428-498d-92fc-1a0a420878d9"
/>
TUI2:
<img width="1089" height="438" alt="image"
src="https://github.com/user-attachments/assets/dd768b1a-2f5f-4bd6-98fd-e52c1d3abd9e"
/>
- Merge ModelFamily into ModelInfo
- Remove logic for adding instructions to apply patch
- Add compaction limit and visible context window to `ModelInfo`
