The Agent Skills Directory

[SAFE]: No security threats or malicious patterns were detected. The skill performs standard web research operations using search engines and page fetching to gather information as documented.\n- [SAFE]: The skill explicitly instructs the agent to respect access boundaries, rate limits, and site terms when browsing the web, which is consistent with benign behavior.\n- [SAFE]: While the skill interacts with untrusted web content (creating a surface for indirect prompt injection), the risk is addressed through a structured workflow that emphasizes evidence-backed synthesis and triangulation across multiple sources. (1) Ingestion points: content fetched via Brave/Bing and webfetch as described in SKILL.md; (2) Boundary markers: implicit via instructions for specific evidence extraction; (3) Capability inventory: search and retrieval tools only, with no code execution or sensitive file access; (4) Sanitization: synthesis and citation-focused processing.

deep-research