finishing-a-development-branch
Pass
Audited by Gen Agent Trust Hub on Sep 27, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes standard command-line tools including
git,gh(GitHub CLI),npm,cargo,pytest, andgo test. These are used for their intended purpose of version control management, repository interaction, and executing local test suites. - [INDIRECT_PROMPT_INJECTION]: The skill processes external data surfaces that could potentially contain untrusted content, such as test runner outputs and git commit histories.
- Ingestion points: External data enters the agent context via test failure logs in Step 1 and commit lists/worktree paths in Step 4.
- Boundary markers: The skill does not use explicit boundary markers or delimiters when presenting test failures or commit lists.
- Capability inventory: Capabilities include shell command execution (git, npm, gh), Pull Request creation, and local file system modifications (deleting branches and worktrees).
- Sanitization: The skill does not perform specific sanitization of test outputs or commit messages before including them in reports or PR bodies, though the use of quoted heredocs (
'EOF') in the PR creation step follows shell security best practices to prevent interpolation. - [SAFE]: The skill demonstrates defensive design by preventing merges when tests fail, requiring explicit string-match confirmation ('discard') before deleting work, and strictly defining a set of four concise options to avoid ambiguous agent behavior.
Audit Metadata