competitor-profiling
Pass
Audited by Gen Agent Trust Hub on Oct 5, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external websites and review platforms, creating a vulnerability surface for instructions hidden in those pages.
- Ingestion points: External competitor URLs and review site URLs (e.g., G2, Capterra) processed during Phase 1 site scraping.
- Boundary markers: The skill includes a specific 'Untrusted Input' core principle that explicitly warns the agent to ignore embedded instructions and treat fetched content as data only.
- Capability inventory: The agent uses web scraping tools (Firecrawl, WebFetch) and SEO data tools (DataForSEO) to read external information, and file system operations to write synthesized profile documents to the local 'competitor-profiles/' directory.
- Sanitization: The instructions require the agent to qualitatively analyze content and report any detected injection attempts rather than executing them.
Audit Metadata