skills/kensaurus/skills/research/Gen Agent Trust Hub

research

Pass

Audited by Gen Agent Trust Hub on Oct 7, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill inherently processes untrusted data fetched from the internet via Firecrawl and Context7 to generate implementation plans.
  • Ingestion points: Data enters the agent's context through firecrawl:firecrawl_search, firecrawl:firecrawl_scrape, firecrawl:firecrawl_map, and context7:query-docs as described in Steps 2, 3, and 5.
  • Boundary markers: The skill mandates a 'Grounding contract' (Step 5b) which requires every fact to be cited with a URL and timestamp. It also enforces a strict 'Output Template' to structure findings separately from inferences.
  • Capability inventory: The skill possesses the ability to read local dependency manifests and codebase files to formulate queries, and it can perform network searches and scraping.
  • Sanitization: The skill includes 'Validation Gates' (Step 6) to reject research patterns that are insecure (CVEs), deprecated, or incompatible with the project's architecture. It also features a 'Pre-Implementation Checklist' to ensure findings are cross-referenced.
  • [EXTERNAL_DOWNLOADS]: The skill fetches external documentation and web content using third-party services.
  • Evidence: Steps 2 and 3 describe the process of resolving library IDs and scraping authoritative URLs (official docs, blogs, engineering reports) to gather technical information.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 7, 2026, 11:27 PM