The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the npx utility to execute the skills CLI tool for searching (find), checking (check), and updating (update) skill packages.
[REMOTE_CODE_EXECUTION]: The npx skills add <package> command downloads and installs modular packages from external sources like GitHub. The instructions specifically recommend using the -y flag, which bypasses confirmation prompts, and the -g flag for global installation, increasing the risk of executing untrusted code without user oversight.
[EXTERNAL_DOWNLOADS]: The skill interacts with the skills.sh registry and various GitHub repositories to fetch skill definitions and executable content.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted metadata from the open skill ecosystem via npx skills find. A malicious package could include instructions in its description or name designed to influence the agent's behavior during the search or installation process.
Ingestion points: Output from the npx skills find [query] command in SKILL.md.
Boundary markers: None explicitly used for search result interpolation.
Capability inventory: Subprocess execution via npx, global package installation, and remote code updates.
Sanitization: The instructions provide manual heuristics (checking stars and install counts) but no automated sanitization of search results.

find-skills