### What version of the Codex App are you using (From “About Codex” dialog)? 26.715.72028 ### What subscription do you have? On the Windows version of Codex, when **Sol** is selected as the main model, **Luna cannot be used as a sub-agent**. I would expect Sol to be able to delegate tasks to Luna as a sub-agent, but Luna either does not appear as an availabl
## Summary The Codex multi-agent sub-agent launcher rejects `gpt-5.6-luna` as an unknown model even though Luna is configured for the Builder role and a direct Codex Luna Max run succeeds on the same machine and ChatGPT account. ## Environment - macOS 26.5.1, arm64 - Codex desktop-bundled CLI: `0.147.0-alpha.6.5` - Authentication: ChatGPT - Global config: `m
### What version of the Codex App are you using (From “About Codex” dialog)? 26.803.10989.0 ### What subscription do you have? Plus ### What platform is your computer? Microsoft Windows NT 10.0.26200.0 x64 ### What issue are you seeing? Codex Desktop fails to start when WSL is selected as the execution environment. <img width="344" height="135" alt="Image" s
Feedback session ID: 019ff564-1d6b-72c1-87e3-87e48b4120fc. desktop build 26.803.10989.0. ChatGPT Business projects visible on chatgpt.com are missing from the Windows desktop app. Account: Devin McGarvey (@devin594646) Workspace: Sift Fire Investigations Desktop build: OpenAI.Codex 26.803.10989.0 Affected project: Sift Technical Writer The account and Busine
### What version of Codex are you using? Codex Desktop package `26.803.10989.0`; source reproduction and candidate fix were made from `openai/codex` main at `b1373b74a27d1d9b65074a873202683355cae772`. ### What platform are you using? Windows 11 x64, elevated Windows sandbox, workspace on a session-scoped mapped drive. ### What issue are you seeing? The eleva
## Summary With macOS Stage Manager enabled, Codex Computer Use can select an inactive app's Stage Manager thumbnail instead of its real window. The thumbnail has off-screen/negative window coordinates, so ScreenCaptureKit fails its display lookup. That failed capture can then poison the shared capture service: subsequent captures of unrelated, valid on-scre
### What issue are you seeing? Recurring Scheduled tasks in ChatGPT Work on the web are sometimes changing from enabled to paused after a scheduled run, without any user request to pause, disable, delete, or reschedule them. In the latest occurrence, four unrelated recurring tasks were found disabled shortly after their runs. Other recurring tasks remained e
### Product Codex Cloud / GitHub `@codex` mention workflow ### Summary A non-review `@codex` mention on an existing pull request started the expected Codex Cloud work task, checked out the correct repository and PR branch, edited and tested the code, and made a local commit. The task was not given any GitHub publishing tool, so it could not push the commit b
## What issue are you seeing? A custom base-instruction override supplied through `model_instructions_file` replaces the model's bundled Codex base instructions in standard Responses mode, but becomes additive in Responses Lite. With Responses Lite enabled, the model sees both: - the service-provided Codex base identity (beginning with `You are Codex`), and
# [Windows] Codex Desktop 26.803.10989.0: "Error: write EOF" main-process modal keeps appearing (with log evidence and my analysis) ## The problem Ever since I updated the Codex Windows desktop app, this modal keeps popping up. It shows on nearly every startup, and then again at random points while I'm working: ``` A JavaScript error occurred in the main pro
### What variant of Codex are you using? CLI ### What feature would you like to see? ### Feature request Codex should support handing off a long-running session from one model provider to another while preserving its working context. Example: - Continue a long GPT session with a custom Responses-compatible provider when the GPT quota is exhausted. - Later ha
### What version of Codex CLI is running? codex-cli 0.146.1 ### What subscription do you have? ? ### Which model were you using? gpt 5.6 luna ### What platform is your computer? cachyos ### What terminal emulator and version are you using (if applicable)? Konsole ### Codex doctor report ```json ~/_newmodel> codex doctor Codex Doctor v0.146.1 · linux-x86_64 N
### What version of the Codex App are you using (From “About Codex” dialog)? 26.803.81509 ### What subscription do you have? ChatGPT Plus ### What platform is your computer? Microsoft Windows NT 10.0.26200.0 x64 ### What issue are you seeing? Browser Use in the Codex Windows desktop app blocks raw JSON and CSV responses with `ERR_BLOCKED_BY_CLIENT`, while or
### What variant of Codex are you using? App Server ### What feature would you like to see? ## Summary ### What we are trying to achieve Context: The `microsoft/vscode` team is integrating the Codex App Server into VS Code's agent host (see our [docs 1](https://code.visualstudio.com/docs/agents/run/agent-harnesses?agent-harness=codex#_use-an-agent-harness) a
### What version of the Codex App are you using (From “About Codex” dialog)? 26.803.81509 ### What subscription do you have? Plus ### What platform is your computer? Linux 7.1.8-100.fc43.x86_64 x86_64 ### What issue are you seeing? When trying to collapse a top-level folder, such as apps, the folder does not collapse immediately. Instead, each click collapse
## Description A long-running Codex Desktop turn was forcibly interrupted while I was asleep. I did not press Stop or provide any input. The local session transcript shows three automated safety rejections of proposed file changes. Immediately after the third rejection, Codex recorded: - Tool result: `aborted by user after 4.2s` - Turn status: `turn_aborted`
### What version of the Codex App are you using (From “About Codex” dialog)? 26.803.10989.0 (Windows; embedded CLI version 0.147.0-alpha.6.6) ### What subscription do you have? ChatGPT Plus ### What platform is your computer? Microsoft Windows NT 10.0.26200.0 x64 ### What issue are you seeing? Two reproducible issues in the Codex App desktop UI: 1. **Popout
### What version of the Codex App are you using (from the About dialog)? OpenAI Codex Windows package `26.803.10989.0` (x64), bundled backend `codex-cli 0.147.0-alpha.6.6`. ### What subscription do you have? ChatGPT sign-in. Exact plan is omitted because the failure occurs in the local Desktop/app-server transport before a model turn starts. ### What platfor
apply_patch "Update"/"Delete" always fails on Windows Desktop: fs sandbox helper stack overflow (0xC0000005 in essctl.dll); "Add" works ## Describe the bug On Codex Desktop for Windows (version 26.803.10989.0), `apply_patch` operations that modify or delete an existing file fail deterministically before applying anything. The tool reports that the filesystem
### What version of the Codex App are you using (From “About Codex” dialog)? 26.803.81509 ### What subscription do you have? Plus ### What platform is your computer? Linux 7.1.8-100.fc43.x86_64 x86_64 ### What issue are you seeing? Links opened from the integrated terminal in the ChatGPT Codex app for Linux are being launched in the system's external browser
### What variant of Codex are you using? App ### What feature would you like to see? Add a keyboard shortcut to quickly comment and uncomment selected lines in the editor, similar to the behavior available in editors such as VS Code. Currently, there does not appear to be a shortcut for toggling line comments in the Codex app editor. This makes a very common
### What variant of Codex are you using? IDE Extension (openai.chatgpt@26.5803.61601) in VS Code 1.133.0 on Linux. ### What feature would you like to see? Please add two independent display controls for edited-file cards: - **Expand edited-file diffs by default** - **Show diff preview when hovering edited-file cards** Both behaviors should be independently c
### What version of the Codex App are you using (From “About Codex” dialog)? 26.803.61601 ### What subscription do you have? ChatGPT Plus ### What platform is your computer? macOS, Apple Silicon (arm64) ### What issue are you seeing? On August 13, 2026, the macOS app returned content apparently belonging to a separate conversation. In an independent non-proj
### What version of the Codex App are you using (From “About Codex” dialog)? OpenAI.Codex MSIX `26.803.10989.0` ### What subscription do you have? Not relevant to this local filesystem behavior. ### What platform is your computer? - Windows, x64 - Codex Desktop agent mode: WSL - WSL2 distribution: Ubuntu 26.04 - WSL kernel: `6.18.33.2-microsoft-standard-WSL2
### What version of Codex CLI is running? `codex-cli 0.147.0` (reported as current/latest by `codex doctor --json`) ### What subscription do you have? ChatGPT authentication (the exact subscription tier is not exposed by the CLI) ### Which model were you using? `gpt-5.6-sol`, max reasoning, fast mode ### What platform is your computer? `Linux 7.0.0-28-generi
### What version of the Codex App are you using (From “About Codex” dialog)? OpenAI.Codex MSIX `26.803.10989.0` ### What subscription do you have? ChatGPT Plus ### What platform is your computer? Windows, x64 ### What issue are you seeing? In Codex Desktop on Windows, the **Choose project** control on the **New chat** screen is visible on screen but is not a
### Feature request Codex Desktop currently starts the configured Git marketplace auto-upgrade as an app-server startup task. When the Desktop app stays open for days, marketplace and plugin updates are not checked again unless the app server is restarted or the user manually requests an upgrade. Current implementation references: - The marketplace auto-upgr
### What version of Codex CLI is running? 0.147.0 ### What subscription do you have? ChatGPT Plus ### Which model were you using? gpt-5.6-sol ### What platform is your computer? Microsoft Windows NT 10.0.26200.0 x64 ### What terminal emulator and version are you using (if applicable)? Windows Terminal with Windows PowerShell 5.1 ### Codex doctor report ```js
### Problem The GitHub-hosted Codex reviewer can reject a requested security review with only: > You have reached your Codex usage limits for security reviews. Please try again later. Example: https://github.com/autumngarage/touchstone/pull/827#issuecomment-5283708031 (2026-08-13). The PR is public and the review was requested against exact head `e9780c11116
This issue was closed and its original details removed at the reporter's request. The problem will be reported through private OpenAI support channels instead.
Fuse recent and lexical skills in shadow selection ## What changed - Add the `lru_plus_lexical_v1` selector, combining the 50 most recent skills with weighted lexical matches through reciprocal rank fusion. - Include the combined selector in the shadow-selection experiment and propagate query and candidate truncation metadata from both inputs. - Generalize t
## Summary - Explicitly protect `.codex/config.toml` as a read-only leaf in split filesystem policies. - Preserve intentional pre-existing project config symlinks. - In bubblewrap, first remount `.codex/` read-only to prevent retargeting, then remount the config symlink's stable target read-only. - Fail closed for dangling config symlinks or targets that tra
Add conservative restriction helpers for tool policies ## What changed - Add `restrict_to` helpers that combine parent and requested `WebSearchMode` and `AppToolApproval` values without granting access beyond either policy. - Treat `Auto` and `Writes` approval modes as incomparable, falling back to `Prompt` when they are combined. ## Testing - Cover every pa
Run search tool integration tests on Windows ## Why The search tool integration test suite was disabled entirely on Windows. ## What changed - Remove the suite-wide Windows exclusion and build test fixtures with automatic environment selection. - Configure stdio MCP fixtures with environment-aware server paths and environment IDs for local and remote runs. -
Stop overriding environments in the skills user-turn test <!-- copyberry-projection-id: cb58cc7482c88b04451c890d241dbf342a8a68bd211ea599fe86a525252be307 -->
Run plugin app-server tests in automatic environments ## What changed - Use the standard automatic test environment for plugin list, install, and uninstall app-server tests. - Keep the bundled stdio MCP test host-local because its fixture is a local executable. - Strengthen that test to verify the bundled server initializes and exposes its `echo` tool. <!--
Add an LRU baseline to skill shadow selection ## What changed - Track the 50 most recently invoked skills for each thread and evaluate them as the `lru_v1` shadow-selection method. - Filter stale and duplicate entries while preserving recency order. - Increase the shadow result limit to 50 and add a `21_50` rank bucket. ## Testing - Add unit coverage for rec
Enforce non-interactive approval policy for Codex delegates ## What changed - Require delegated Codex sessions to use the `never` approval policy and reject delegate creation with any prompt-capable policy. - Stop forwarding delegate approval and permission requests to the parent session. Approval-requiring commands and MCP tool calls are denied within the d
Avoid allocations when sanitizing TUI user text ## Why Sanitizing user messages and pasted text always allocated a new string, even when the input contained no terminal control characters. ## What changed - Return a `Cow<str>` from `sanitize_user_text` so clean borrowed input remains borrowed. - Reuse owned input buffers when sanitization leaves the text int
Add a flag to retain client developer messages ## What changed - Register `retain_client_developer_messages` as an under-development feature that is disabled by default. - Expose the feature in the configuration schema. <!-- copyberry-projection-id: 29b030fa11a45d0b8aefa5cbf378976407e2b239d74721bf86efdbd641812823 -->
Add manifest-defined metrics for trusted plugin scripts ## What changed - Load version 1 `analytics.yaml` manifests from trusted plugin roots and map declared operations, measurements, and enum dimensions to exact script paths. - Expose resolution types and bind a declared operation to the plugin identity returned by fresh command attribution. - Reject malfo
Add bounded plugin measurement analytics ## What changed - Add `PluginMeasurementsInput` and `PluginMeasurementRow`, plus an analytics client method for recording batches and checking whether analytics are enabled. - Emit each valid row as a `codex_plugin_measurement_event` with thread, turn, item, plugin, execution, and operation context. - Bound batches at
Cache stable active-cell layout measurements ## What changed - Reuse desired and rendered heights for active transcript cells that report a stable height. - Invalidate the cached layout when the cell identity, revision, width, render mode, or syntax theme changes, and when the active cell is removed. - Continue rebuilding display lines on every frame so anim
Add per-thread usage queries to the backend client ## What changed - Add `Client::get_thread_usage` for querying authoritative estimated credit and dollar usage for a thread across both supported backend path styles. - Expose usage totals and optional model, reasoning effort, speed, and token breakdowns through `ThreadUsage` and `ThreadUsageBreakdownGroup`.
Represent persisted world state as JSON objects ## Why World-state snapshots and merge patches are keyed collections of sections. Allowing their persisted `state` field to contain any JSON value required replay code to handle shapes that cannot represent world state. ## What changed - Type `WorldStateItem.state` and snapshot patches as JSON object maps. - Re
Use protobuf's built-in Bazel proto rule ## What changed - Load `proto_library` for the code-mode protocol directly from the `com_google_protobuf` repository. - Remove the now-unneeded direct `rules_proto` module dependency. <!-- copyberry-projection-id: e6b84803e28824c3b7e0593badf3b7920e0c913bb09709b7f2b30475558c485f -->
Move `codex-execpolicy` to protocol dev dependencies ## What changed Declare `codex-execpolicy` as a dev dependency of `codex-protocol`, matching its use exclusively by tests. <!-- copyberry-projection-id: 81f7b680437dc90a9e7f763db2b620f05f2154c43e5cab3aeaaef43e73e9a9e1 -->
Remove unused apply_patch prompt fallback ## What changed - Delete the unused standalone `apply_patch` instruction template and its `codex-prompts` export. - Simplify prompt-caching coverage to assert that requests use the model's base instructions. <!-- copyberry-projection-id: 3d9409b8d9b35e31e06739a14b824035904bc928b371dae2c258b7f4e97b3083 -->
Add Guardian V2 extension scaffold ## What changed - Add the `codex-guardian-v2` crate with an extension install entry point that does not register contributors yet. - Register the crate in the Cargo workspace and add its Bazel target. <!-- copyberry-projection-id: c87cd1228d948473717d12eeada1529d3653abc088ae33010ff77f15cf03ff09 -->
Optimize orphan output normalization ## What changed - Collect borrowed call IDs in a single pass instead of cloning IDs across separate sets. - Record orphan positions and compact the history only when orphan outputs are present, while preserving the existing matching and error behavior. <!-- copyberry-projection-id: af969a7d4fa2b7defb08f3a5c0daa10c15b712bd