Harden generated DISPATCH_TOKEN workflows

[jeffhuber]

Summary

• run generated workflows that receive DISPATCH_TOKEN from trusted default-branch definitions via pull_request_target
• retain the same-repository guard and avoid checking out pull-request code
• cover both source and packaged templates with focused security regression tests

Verification

• full unittest suite: 853 passed
• focused init and release-hygiene tests passed
• Ruff, privacy scan, compileall, and easy-mode smoke test passed

Provenance

• builder: Antigravity CLI
• finding source: Claude audit of jeffhuber/codemower.com#171

Closes #654

[gitar-bot]

Code Review ✅ Approved

Hardens generated DISPATCH_TOKEN workflows to run from trusted default-branch definitions via pull_request_target, retaining same-repository guards and avoiding checkout of pull-request code. Security regression tests added for both source and packaged templates. No issues found.

Options

Auto-apply is off → Gitar will not commit updates to this branch.
Display: compact → Showing less information.

Comment with these commands to change the behavior for this request:

Auto-apply	Compact
gitar auto-apply:on	gitar display:verbose

_{Was this helpful? React with 👍 / 👎 | Gitar}

[jeffhuber]

Codex audit (merge-authority lane)

Head SHA: a6d04601cdb6982498adcc15490d0056eabd25a1
Findings: P0=0, P1=0, P2=0, P3=0 (blocker policy: any P0/P1/P2 → BLOCKED)

Codex Audit: PASS

Summary:

No blocking issues were found.

Findings: none.

[jeffhuber]

Claude audit (merge-authority lane)

Head SHA: a6d04601cdb6982498adcc15490d0056eabd25a1
Findings: P0=0, P1=0, P2=0, P3=0 (blocker policy: any P0/P1/P2 -> BLOCKED)

Claude Audit: PASS

Summary:

Switching the agent-PR-labeler and fix-round-dispatch templates from pull_request to pull_request_target is a legitimate hardening: it pins execution to the base-branch workflow definition (so a same-repo PR branch can't tamper with the workflow steps to exfiltrate DISPATCH_TOKEN), no PR code is checked out or executed, the existing same-repo if guard is preserved, and both template copies plus new tests are updated consistently.

Findings: none.