tool-fit-for-projects
Pass
Audited by Gen Agent Trust Hub on Nov 2, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill authorizes the agent to execute existing tests, benchmarks, and diagnostic commands found within the current repository to validate tool fit.
- Evidence: Found in SKILL.md under 'Build the Evidence Base', which states: 'Read-only Validation may run existing tests, benchmarks, and diagnostic commands that require no dependency, source, configuration, or service mutation.' This is a feature of the tool audit process and is constrained to existing repository code.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from the repository being audited, including manifests, code comments, and historical issues, which could potentially contain malicious instructions.
- Ingestion points: SKILL.md ('Build the Evidence Base' section) specifies reading repository code, manifests, configuration, history, and issues.
- Boundary markers: The skill provides instructions to 'Trace every material claim' and 'Separate verified facts, repository-specific inference, user constraints, and unknowns', which act as a logical buffer against accepting embedded claims at face value.
- Capability inventory: The agent has the ability to read all files in the current directory and execute existing diagnostic scripts/tests.
- Sanitization: No explicit prompt sanitization is mentioned, but the skill emphasizes verifying claims against multiple levels of an evidence hierarchy.
Audit Metadata