autopilot
Pass
Audited by Gen Agent Trust Hub on Aug 25, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION]: The skill's primary function involves surveying a codebase (README, code comments, commit history) and acting on its contents. This establishes a surface for indirect prompt injection where malicious instructions embedded in the codebase could attempt to influence the agent's autonomous decisions (found in SKILL.md).\n
- Ingestion points: Codebase survey in Step 1 (README, CLAUDE.md, source code, commit history).\n
- Boundary markers: The skill includes a comprehensive 'Stay in bounds' section and 'Surface first' rules that explicitly prevent the agent from performing high-risk actions (touching secrets, auth, or deployments) without human review.\n
- Capability inventory: The skill leverages the agent's standard capabilities for reading/writing files, running tests, and managing version control.\n
- Sanitization: No specific sanitization or escaping is defined for the content read from the codebase, though the logical constraints act as a behavioral safeguard.
Audit Metadata