Codex desktop should detect newer parallel branches before preparing releases

Product feedback

When several ChatGPT/Codex tasks work on the same GitHub repository, Codex Desktop can prepare a technically valid release from main while newer, relevant design work is still present on another active branch or pull request. In this case, the agent fetched GitHub and noticed the newer branch, but still based its release on main. Automated quality and security checks passed, yet the resulting release was functionally incomplete and omitted the parallel design changes.

User impact

• Previously completed design work was missing from the release candidate.
• Successful CI checks created false confidence because the source baseline was incomplete.
• The user had to detect the mistake and repeat verification and recovery work.
• Repeated permission prompts further interrupted the workflow.
• Codex Light consumed credits and took longer while creating more supervision work than ChatGPT High.

Suggested improvements

1. Before changing, packaging, or releasing a repository, inspect active branches and open pull requests for newer relevant work, not only main.
2. Warn clearly when another branch changes overlapping files or contains a newer candidate state.
3. Require explicit baseline confirmation when the newest relevant state is ambiguous.
4. Distinguish technical CI success from completeness against parallel work.
5. Bundle related Git/GitHub permission requests and reuse narrowly scoped approvals wherever safe.
6. Keep every project adjustment traceable on the active branch, including review-stage assets, instead of leaving changes only outside the repository.

No private repository contents, business data, credentials, or customer information are included in this report.

No files, tests, or entry points are named. Start by locating the release-preparation workflow and its GitHub branch and pull-request discovery logic, then determine how relevant newer work and ambiguous baselines should be identified. Done should include a clear implementation scope and tests for incomplete parallel-work detection.