commit-work
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from the repository's file system, which could potentially contain malicious instructions intended to influence the agent's behavior during the commit process.
- Ingestion points: The skill reads
git status, unstaged diffs, staged diffs, and repository instructions (SKILL.md, Workflow steps 1 and 6). - Boundary markers: There are no explicit instructions or delimiters mentioned to distinguish between user-provided repository instructions and potentially malicious content within the files being diffed.
- Capability inventory: The agent can execute Git commands (
add,commit) and verification tasks (SKILL.md, Workflow steps 5, 7, and 9). - Sanitization: No specific sanitization or filtering of the file contents or diff output is described before the agent processes them.
- [DYNAMIC_EXECUTION]: The skill instructs the agent to execute verification checks which involve running local code or scripts found within the repository.
- Evidence:
SKILL.md(Workflow step 7) directs the agent to "Run the smallest meaningful verification for the staged behavior". This requires executing test runners or validation scripts (e.g.,npm test,pytest) that are dynamically determined by the environment, representing a surface for executing code present in the worktree.
Audit Metadata