[ca] Update Codex CLI to version 0.78.0

[github-actions]

Version Update: Codex CLI 0.77.0 → 0.78.0

Previous Version: 0.77.0 (released December 21, 2025)
New Version: 0.78.0 (released January 6, 2026)
Release Timeline: ~16 days between releases
Commits in this release: 20 commits

---

Change Summary

Breaking Changes

None detected

New Features

• Firewall Rules API Enhancement: Added justification argument to prefix_rule() in *.rules files
• App Server API: Exposed outputSchema to user_turn/turn_start app server API endpoints
• TUI2 Scrollbar: New transcript scrollbar with auto-hide and drag functionality
• MacOS MDM Support: Added ability to load config requirements from MDM on macOS

Bug Fixes

• Sandbox Security: Fixed sandbox to never allow writes to .codex/ or .codex/.sandbox/ directories
• Firewall Setup: Improved idempotency for creating/updating firewall rules during setup
• Chat Tool Calls: Fixed handling of multiple tool calls in chat mode
• Review Command: Fixed /review command to respect session current working directory (cwd)
• Codex API: Fixed Chat Completions DONE sentinel handling
• Device Auth: Updated device auth prompt to use issuer URL

Security Improvements

• Sandbox User Privacy: Best effort implementation to "hide" sandbox users from system
• Sandbox Write Protection: Hardened protection against writes to .codex/ directories

Performance & Quality

• MCP Tool Names: Sanitized MCP tool names to ensure compatibility with Responses API
• GitHub Pager: Improved GitHub paging functionality
• Skills Rendering: Enhanced skills render section display
• TUI2 Selection: Copy selection now dismisses highlight for better UX

Dependency Updates

• insta: 1.44.3 → 1.46.0
• derive_more: 2.0.1 → 2.1.1
• tokio-stream: 0.1.17 → 0.1.18
• clap_complete: 4.5.57 → 4.5.64

---

CLI Changes Analysis

Result: No CLI interface changes detected between versions 0.77.0 and 0.78.0

The --help output for both the main command and all subcommands remains identical. All command-line flags, options, and arguments are unchanged.

---

Impact Assessment

Risk Level: Low

Affected Features:

• Sandbox security (improved protection)
• Firewall setup reliability
• TUI2 user experience enhancements
• MCP server compatibility

Migration Required: No

This is a maintenance release focused on bug fixes, security hardening, and incremental improvements. No breaking changes or significant behavioral changes that would affect gh-aw integration.

---

Merged Pull Requests

Key PRs included in this release:

1. Security & Sandbox:

   • never let sandbox write to .codex/ or .codex/.sandbox/ openai/codex#8683 - Prevent sandbox writes to .codex/ directories
   • best effort to "hide" Sandbox users openai/codex#8492 - Hide sandbox users from system
   • better idempotency for creating/updating firewall rules during setup. openai/codex#8686 - Improve firewall rule idempotency

2. Features:

   • feat: add justification arg to prefix_rule() in *.rules openai/codex#8751 - Add justification to prefix_rule() API
   • feat: expose outputSchema to user_turn/turn_start app_server API openai/codex#8377 - Expose outputSchema in app server API
   • feat(tui2): transcript scrollbar (auto-hide + drag) openai/codex#8728 - TUI2 scrollbar with auto-hide & drag
   • (MacOS) Load config requirements from MDM openai/codex#8743 - MacOS MDM config support

3. Bug Fixes:

   • fix: /review to respect session cwd openai/codex#8738 - Fix /review to respect session cwd
   • fix: chat multiple tool calls openai/codex#8556 - Fix chat multiple tool calls
   • fix(codex-api): handle Chat Completions DONE sentinel openai/codex#8708 - Fix Chat Completions DONE sentinel
   • Use issuer URL in device auth prompt link openai/codex#7858 - Use issuer URL in device auth

4. Quality Improvements:

   • [MCP] Sanitize MCP tool names to ensure they are compatible with the Responses APO openai/codex#8694 - Sanitize MCP tool names for Responses API
   • chore: GH pager openai/codex#8747 - Improve GH pager
   • chore: improve skills render section openai/codex#8459 - Improve skills render section
   • tui2: copy selection dismisses highlight openai/codex#8718 - TUI2 copy selection UX

5. Dependencies:

   • chore(deps): bump insta from 1.44.3 to 1.46.0 in /codex-rs openai/codex#8725 - Bump insta to 1.46.0
   • chore(deps): bump derive_more from 2.0.1 to 2.1.1 in /codex-rs openai/codex#8724 - Bump derive_more to 2.1.1
   • chore(deps): bump tokio-stream from 0.1.17 to 0.1.18 in /codex-rs openai/codex#8723 - Bump tokio-stream to 0.1.18
   • chore(deps): bump clap_complete from 4.5.57 to 4.5.64 in /codex-rs openai/codex#8721 - Bump clap_complete to 4.5.64

---

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-v0.78.0

---

Update Status

✅ Updated: pkg/constants/constants.go - Changed DefaultCodexVersion from "0.77.0" to "0.78.0"
✅ Recompiled: 53 workflow files updated via make recompile
✅ Verified: All changes applied successfully

Files Modified: 53 (1 source file + 52 workflow lock files)
Lines Changed: +99, -81

---

Testing Recommendations

1. Test Codex agent execution in gh-aw workflows
2. Verify sandbox security restrictions are working correctly
3. Confirm MCP server integration remains functional
4. Test firewall rule creation/updates in activation workflows

---

Notes

This update is part of the regular CLI version maintenance cycle. The changes are primarily internal improvements that should not impact gh-aw's use of Codex. The security hardening around sandbox write restrictions and improved firewall setup reliability are positive additions that enhance overall system security.

AI generated by CLI Version Checker