The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill facilitates the ingestion of untrusted data from the open web through search and extraction tools, creating a vulnerability surface where external content can influence agent behavior.
Ingestion points: Web content retrieved via mcp__tavily__tavily_search and mcp__tavily__tavily-extract (SKILL.md).
Boundary markers: The instructions lack specific guidance on using delimiters or instructions to ignore embedded commands in the retrieved data.
Capability inventory: The skill is intended to be used by an agent with broader capabilities (like shell or file access), which could be targeted by malicious instructions embedded in scraped websites.
Sanitization: There is no mention of sanitizing, filtering, or validating the content extracted from external URLs before the agent processes it.
[COMMAND_EXECUTION]: The skill documentation references a local script tavily_extract_to_advanced.py described as a hook that automatically modifies tool calls. Execution of local scripts represents a capability that should be monitored.

tavily-usage