redesign-existing-projects
Pass
Audited by Gen Agent Trust Hub on Oct 30, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from existing codebases which may contain malicious instructions.
- Ingestion points: The skill performs a scan of the local codebase to identify frameworks and styling patterns (SKILL.md).
- Boundary markers: No specific delimiters or instructions are provided to distinguish between scanned code and agent instructions to prevent accidental obedience to embedded strings.
- Capability inventory: The agent has permission to modify project files, update styling systems, and check dependency files (SKILL.md).
- Sanitization: There is no explicit process to sanitize or filter input from the scanned codebase to prevent the execution of hidden prompts.
- [EXTERNAL_DOWNLOADS]: Suggests using the well-known picsum.photos service to fetch placeholder images for design audits when real assets are unavailable (SKILL.md).
Audit Metadata