[ca] Update OpenAI Codex CLI to v0.95.0

[github-actions]

Updated OpenAI Codex CLI from v0.94.0 to v0.95.0, including new features for macOS app integration, personal skill loading, and parallel shell execution.

Update Summary

• Previous Version: 0.94.0 (February 2, 2026)
• New Version: 0.95.0 (February 4, 2026)
• Release Timeline: 2 days between releases
• Breaking Changes: No
• Update Priority: Medium

Key Features

macOS App Launcher

• New codex app (path) command to launch Codex Desktop from CLI with automatic DMG downloads
• Improves integration between CLI and desktop versions

Personal Skill Loading

• Support for loading skills from ~/.agents/skills (with ~/.codex/skills compatibility)
• App-server APIs to list and download public remote skills
• Enables better customization and extensibility

Enhanced /plan Command

• Now accepts inline prompt arguments and pasted images
• Improved slash-command editing/highlighting in the TUI
• Better user experience for planning workflows

Parallel Shell Execution

• Shell-related tools can now run concurrently
• Improves multi-command throughput and performance

Thread ID Environment Variable

• Shell executions now receive CODEX_THREAD_ID
• Enables script/skill detection of active sessions

CLI Changes

View CLI Changes

New Command Structure:

• sandbox command no longer has [aliases: debug]
• debug is now a separate standalone command (no longer an alias)
  • New dedicated "Debugging tools" command

Local Provider Changes:

• --local-provider flag options updated
  • Removed: ollama-chat option
  • Remaining: lmstudio and ollama only

Bug Fixes

View Bug Fixes

Git Safety

• Hardened command safety to prevent destructive/write-capable invocations from bypassing approval checks
• Related PR: fix: unsafe auto-approval of git commands openai/codex#10258

Thread Management

• Fixed saved thread names display and thread listing behavior
• Improves resume/browsing reliability
• Related PRs: Session picker shows thread_name if set openai/codex#10340, fix: thread listing openai/codex#10383

Sandbox Mode

• Fixed first-run trust-mode reporting consistency
• Made $PWD/.agents read-only for better security
• Related PRs: Fixed sandbox mode inconsistency if untrusted is selected openai/codex#10415, fix: make $PWD/.agents read-only like $PWD/.codex openai/codex#10524

Exec Hanging

• Resolved codex exec hanging after interrupt in websocket/streaming scenarios
• Implements clean shutdown
• Related PR: Handle exec shutdown on Interrupt (fixes immortal codex exec with websockets) openai/codex#10519

Approval Events

• Fixed review-mode approval event wiring for proper requestApproval ID alignment
• Related PR: fix(app-server): fix approval events in review mode openai/codex#10416

Error Diagnostics

• Enhanced 401 error messages with server details, cf-ray, and requestId
• Better debugging information for authentication issues
• Related PR: Add more detail to 401 error openai/codex#10508

Merged PRs

View Significant Merged PRs

Features:

• Add codex app macOS launcher openai/codex#10418 - macOS App Launcher
• feat: Read personal skills from .agents/skills openai/codex#10437 - Personal skill loading from ~/.agents/skills
• feat: add APIs to list and download public remote skills openai/codex#10448 - App-server APIs for remote skills
• Nicer highlighting of slash commands, /plan accepts prompt args and pasted images openai/codex#10269 - Enhanced /plan command with inline args
• Enable parallel shell tools openai/codex#10505 - Parallel shell execution
• Inject CODEX_THREAD_ID into the terminal environment openai/codex#10096 - CODEX_THREAD_ID environment variable
• feat(linux-sandbox): vendor bubblewrap and wire it with FFI openai/codex#10413 - Linux sandbox enhancement (Bubblewrap)

Bug Fixes:

• fix: unsafe auto-approval of git commands openai/codex#10258 - Git command safety hardening
• Session picker shows thread_name if set openai/codex#10340 - Thread names display fix
• fix: thread listing openai/codex#10383 - Thread listing behavior fix
• Fixed sandbox mode inconsistency if untrusted is selected openai/codex#10415 - Sandbox trust-mode reporting
• fix: make $PWD/.agents read-only like $PWD/.codex openai/codex#10524 - .agents directory read-only
• Handle exec shutdown on Interrupt (fixes immortal codex exec with websockets) openai/codex#10519 - Exec hanging fix
• fix(app-server): fix approval events in review mode openai/codex#10416 - Approval events wiring
• Add more detail to 401 error openai/codex#10508 - Enhanced 401 error messages

Documentation & Maintenance:

• Nicer highlighting of slash commands, /plan accepts prompt args and pasted images openai/codex#10269 - TUI chat composer documentation
• Updated labeler workflow prompt to include "app" label openai/codex#10411 - Issue templates update
• Updated bug and feature templates openai/codex#10453 - Labeler prompts refresh
• feat: add MCP protocol types and rmcp adapters openai/codex#10356 - MCP protocol types migration
• feat: replace custom mcp-types crate with equivalents from rmcp openai/codex#10349 - MCP protocol types migration
• chore: remove deprecated mcp-types crate openai/codex#10357 - MCP protocol types migration
• chore: update bytes crate in response to security advisory openai/codex#10525 - bytes dependency security update

Impact Assessment

Risk Level: Low

• No breaking changes
• Primarily feature additions and bug fixes
• CLI changes are non-breaking (debug becomes separate command, local provider options refined)

Affected Features:

• macOS users benefit from new app launcher
• Users with custom skills benefit from ~/.agents/skills support
• Parallel shell execution improves performance for multi-command workflows
• Enhanced error messages improve debugging experience

Recommendations

Update Priority: Medium

• No critical security vulnerabilities
• Includes useful features and bug fixes
• Git safety hardening is particularly valuable for automated workflows
• Parallel shell execution may improve workflow performance

Testing Strategy:

1. Verify basic CLI functionality works as expected
2. Test new debug command separately from sandbox
3. Validate --local-provider flag with lmstudio and ollama options
4. Monitor workflow execution for any performance improvements from parallel execution

Rollout Plan:

• Update constants.go (completed: pkg/constants/constants.go:272)
• Run make recompile to regenerate workflow files (pending due to network issues)
• Test in staging environment
• Monitor workflow runs for any issues

Package Links

• NPM Package: https://www.npmjs.com/package/`@openai/codex`
• Repository: https://github.com/openai/codex
• Release Notes: https://github.com/openai/codex/releases
• Specific Release: https://github.com/openai/codex/releases/tag/rust-v0.95.0

References

• §21671282571 (Current workflow run)

---

Note: The make recompile step encountered temporary network issues with the Go proxy during this workflow run. This does not affect the version update itself and can be resolved by rerunning the build in a stable network environment.

AI generated by CLI Version Checker

• expires on Feb 6, 2026, 12:28 PM UTC

[github-actions]

🍪 Issue Monster has assigned this to Copilot!

I've identified this issue as a good candidate for automated resolution and assigned it to the Copilot agent.

The Copilot agent will analyze the issue and create a pull request with the fix.

Om nom nom! 🍪

AI generated by Issue Monster

[github-actions]

This issue was automatically closed because it expired on 2026-02-06T12:28:32.095Z.