web-research

Pass

Audited by Gen Agent Trust Hub on Aug 28, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from the internet using WebFetch, WebSearch, and platform-specific CLI tools, creating a surface where malicious instructions embedded in web pages could influence the agent's behavior.
  • Ingestion points: External content fetched from URLs during the '核验' (Verification) and '归档' (Archive) phases as described in SKILL.md.
  • Boundary markers: The skill requires metadata headers (URL, time, method) in archived files but does not specify explicit delimiters or "ignore instructions" wrappers for the content during processing.
  • Capability inventory: The skill utilizes Bash (to run yt-dlp and gh), Write (to save files to the local system), and WebFetch/WebSearch (network operations), as defined in the allowed-tools section of SKILL.md.
  • Sanitization: The workflow relies on manual user confirmation and verification of candidates to mitigate risks, rather than automated programmatic sanitization of the fetched text content.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 28, 2026, 10:08 AM
Security Audit — agent-trust-hub — web-research