Skills
Navigate
skills/practicalswan/agent-skills/tavily-cli/Gen Agent Trust Hub

tavily-cli

Pass

Audited by Gen Agent Trust Hub on Sep 9, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted web content retrieved through the Tavily CLI tools.
  • Ingestion points: The agent ingests data from tvly search, extract, map, crawl, and research commands (SKILL.md).
  • Boundary markers: The skill explicitly warns in the 'Anti-Patterns' and 'Verification Protocol' sections that external content must be treated as untrusted and should not be interpreted as instructions (SKILL.md).
  • Capability inventory: The skill uses the tvly command-line tool, supports file writing via the -o flag, and includes package installation via uv or pip (SKILL.md).
  • Sanitization: Instructions emphasize manual inspection of scripts and using bounded scopes for automated crawling (SKILL.md).
  • [EXTERNAL_DOWNLOADS]: The documentation references the official Tavily CLI installation script.
  • Evidence: Mentions https://cli.tavily.com/install.sh for installation but provides a security warning to inspect the script and avoid piping it directly to the shell (SKILL.md).
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 9, 2026, 03:52 AM