website-reference-curator

Pass

Audited by Gen Agent Trust Hub on Sep 9, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill uses Python's standard library (pathlib.Path.resolve) in scripts/catalog.py to validate that evidence paths stay within the skill's root directory, preventing directory traversal.
  • [SAFE]: Instructions in references/ingestion.md explicitly state that content from external websites must be treated as untrusted data and not as agent instructions, mitigating indirect prompt injection risks.
  • [SAFE]: External references to Awwwards and GitHub repositories for component suppliers are relevant to the skill's purpose and originate from well-known services.
  • [SAFE]: The skill uses SHA-256 checksums to verify the integrity of local screenshot evidence, ensuring data consistency and detecting tampering.
  • [SAFE]: The scripts/catalog.py script implements specific protections against CSV injection (formula execution) when exporting data by prefixing potentially dangerous cells with a single quote.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 9, 2026, 01:03 PM
Security Audit — agent-trust-hub — website-reference-curator