fix(jobs): run agents with unattended permissions - #149
Conversation
|
@codex review |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 88da1e49d6
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
|
@codex review |
|
Codex Review: Didn't find any major issues. Chef's kiss. Reviewed commit: ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
If Codex has suggestions, it will comment; otherwise it will react with 👍. Codex can also answer questions or update the PR. Try commenting "@codex address that feedback". |
…d-job-permissions fix(jobs): run agents with unattended permissions
What changed
danger-full-accessand approval policyneverbypassPermissionsis_error: falsev0.8.1Root cause
Scheduled and manual jobs reused the normal chat runner, which delegated to backend permission defaults. In a headless run, required approval or sandbox access could not be granted. Claude output parsing also trusted a non-error JSON envelope without first enforcing the process exit status.
Impact
Codex and Claude jobs can now complete without an interactive operator. Chats retain their existing permission behavior. Job definitions execute with the OS permissions of the Push service user.
Verification
cargo fmt --all --checkcargo clippy --locked --all-targets -- -D warningscargo build --lockedcargo test --lockedbash tests/install.shbash tests/release-version.shbash scripts/check-release-version.sh v0.8.1mkdocs build --strict