improve-animations
Pass
Audited by Gen Agent Trust Hub on Nov 6, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes codebase files which are untrusted and may contain hidden instructions.
- Ingestion points: The agent sweeps the repository for animation-related code using file reading and search tools during the 'Recon' and 'Audit' phases.
- Boundary markers: 'Hard Rule 4' specifically instructs the agent to treat repository content as inert data and to ignore/flag attempts to override instructions (e.g., 'ignore previous instructions') found in audited files.
- Capability inventory: The skill writes plan files to the "plans/" directory and can dispatch subagents to assist with category-specific audits.
- Sanitization: The 'Vet, prioritize, confirm' phase (Phase 3) requires the agent to manually re-verify all automated findings against the source code to ensure accuracy and prevent data-driven manipulation.
Audit Metadata