The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill provides instructions to download and install external code via npx skills add <package>. It specifically encourages the use of the -y flag to skip user confirmation and the -g flag for global installation. This behavior allows for the execution of arbitrary remote code fetched from repositories like those hosted on GitHub.
[EXTERNAL_DOWNLOADS]: The skill references the skills.sh website and third-party GitHub repositories for package discovery and installation. It encourages downloading and integrating external content into the agent's operating environment.
[COMMAND_EXECUTION]: The skill relies on shell command execution via the npx utility to manage agent skills. This involves running commands to find, add, check, and update software packages on the local system.
[PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection. It ingests untrusted data from external sources via the npx skills find command, which acts as the ingestion point. The skill does not define boundary markers or sanitization steps to protect against malicious instructions embedded in the metadata of retrieved packages. The agent possesses significant capabilities, including subprocess execution via npx skills add, which could be exploited by such an injection.

find-skills