The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes shell functions (exa_rpc, exa_init, exa_call) that use curl, jq, awk, and sed to interact with the Exa MCP endpoint. These commands are necessary for the skill's function of communicating with the search API and parsing server-sent events.
[EXTERNAL_DOWNLOADS]: The documentation references specific versions and SHA-256 integrity hashes for the exa-mcp-server NPM package and its corresponding GitHub repository. These references target a well-known service provider and demonstrate best practices for verifying external dependencies.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection as it ingests data from the public web.
Ingestion points: External content is retrieved via the web_search_exa and web_fetch_exa tools as described in SKILL.md.
Boundary markers: The "Research Practice" section instructs the agent to filter and summarize results rather than providing raw data, which acts as a basic oversight mechanism.
Capability inventory: The skill uses network communication to the Exa API and local shell utilities for data transformation.
Sanitization: There is no explicit description of structural sanitization or safety filtering for the content returned by the search tools.

exa-search