The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill includes setup instructions in references/setup.md that direct the agent or user to install the tvly CLI using a shell pipe pattern: curl -fsSL https://cli.tavily.com/install.sh | bash. While this originates from the vendor's official domain, this execution pattern is a high-risk method for remote code execution.
[EXTERNAL_DOWNLOADS]: The skill instructs on the installation of external dependencies and software:
Installation of the tavily-cli package via pip or uv in references/setup.md.
Use of npx skills add to dynamically install additional skills such as tavily-research from the tavily-ai/skills repository.
[INDIRECT_PROMPT_INJECTION]: The skill's primary function is to search and crawl the web, creating a surface for indirect prompt injection from untrusted external data.
Ingestion points: External web content and search snippets retrieved via the tavily_search, tavily_extract, tavily_map, and tavily_crawl tools.
Boundary markers: The instructions do not define clear delimiters or provide 'ignore instructions' warnings for the agent when processing retrieved web content.
Capability inventory: The skill has the ability to write files to disk (-o <file>), execute CLI commands, and install additional software/skills via shell and package managers.
Sanitization: There is no mention of sanitization, filtering, or validation steps for the data retrieved from external URLs before it is processed by the agent.

tavily-search