research-to-plan

Pass

Audited by Gen Agent Trust Hub on Mar 7, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [SAFE]: The skill uses trusted tools (Firecrawl) and references well-known documentation domains (e.g., react.dev, nextjs.org) for its research phase. Orchestration is handled through standard agent team and task management operations.\n- [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by ingesting external data from documentation websites into a knowledge base.\n
  • Ingestion points: The firecrawl:crawl and firecrawl:map tools in SKILL.md fetch content from external URLs.\n
  • Boundary markers: No specific delimiters or instructions to ignore embedded commands within the crawled content are defined.\n
  • Capability inventory: The skill can perform file system writes (saving design and implementation plans to docs/plans/) and manage agent teams and tasks.\n
  • Sanitization: No explicit sanitization or filtering of the content retrieved via Firecrawl is specified before it is used for design augmentation.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 7, 2026, 05:32 PM