research
Pass
Audited by Gen Agent Trust Hub on Sep 17, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests external data from the Tavily Search API, creating a standard surface for indirect prompt injection common to all research tools.
- Ingestion points: External web content fetched via the Tavily API in
scripts/tavily-cli.ts. - Boundary markers: The research methodology in
SKILL.mdutilizes structured synthesis templates and checklists which help the agent distinguish between search data and final analysis. - Capability inventory: The skill utilizes network access (Tavily API) and outputs results to the terminal.
- Sanitization: Search results are truncated for basic display, though raw content can be processed when requested.
- [SAFE]: The skill performs expected network operations to
api.tavily.com, an established search service, which is necessary for the skill's stated purpose.
Audit Metadata