The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the npx skills command-line utility to search for, install, and update agent capabilities.
[EXTERNAL_DOWNLOADS]: The skill interacts with the skills.sh registry and GitHub to download third-party extensions. The instructions highlight established organizations as examples of reputable sources.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and summarize content from the skills ecosystem, such as search results and skill descriptions, which could theoretically contain adversarial instructions.
Ingestion points: Search output from the CLI and the content of SKILL.md files from newly installed packages.
Boundary markers: The instructions require the agent to analyze and report the installed skill's functionality to the user, acting as a manual review step.
Capability inventory: Includes the ability to install code via npx skills add and execute searches via npx skills find.
Sanitization: The skill relies on reputation-based filtering (author verification, star counts, install volume) and mandatory human approval as the primary security controls.

find-skills