The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes content from untrusted external URLs. * Ingestion points: tvly extract command in SKILL.md is used to ingest data from user-specified URLs. * Boundary markers: The skill contains clear warnings to treat returned content as untrusted and identifies treating external content as trusted instructions as an anti-pattern. * Capability inventory: The skill executes tvly CLI for network operations and supports file-writing with the -o flag as seen in SKILL.md. * Sanitization: Content is passed to the agent without explicit sanitization steps mentioned in the skill.
[COMMAND_EXECUTION]: The skill utilizes the tvly command-line utility. * Evidence: Examples in SKILL.md show the use of tvly extract with various options including --json and -o.

tavily-extract