watermark
Pass
Audited by Gen Agent Trust Hub on Oct 13, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill analyzes uncommitted work via git diff, which allows potentially malicious instructions embedded in source code or comments to influence the agent's behavior during commit generation. 1. Ingestion points: SKILL.md (Step 1) uses git diff to inspect changes. 2. Boundary markers: The instructions lack explicit directives for the agent to ignore or sanitize natural language instructions found within the code changes. 3. Capability inventory: The skill performs write operations using git add, git apply, and git commit (SKILL.md Step 5). 4. Sanitization: Step 3 includes a safety gate that explicitly refuses to stage sensitive files such as .env or credentials.
- [COMMAND_EXECUTION]: The skill uses shell commands to interact with the local git repository. 1. Evidence: SKILL.md invokes git status, git diff, git log, git add, git apply, and git commit to manage the workflow.
Audit Metadata