The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to use npx skills commands to search, check, and update packages. This involves executing shell commands with arguments derived from user queries or remote package data.
[EXTERNAL_DOWNLOADS]: The npx skills add <package> command fetches external content from repositories. The skill specifically directs the agent to use the -y flag, which skips confirmation prompts, effectively automating the download of third-party content without user oversight.
[REMOTE_CODE_EXECUTION]: By facilitating the installation of third-party packages from arbitrary GitHub repositories or sources, the skill enables remote code execution. The instructions promote the use of the global flag (-g), which can result in system-wide installation of potentially untrusted scripts.
[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect injection as it ingests and processes descriptions and metadata from the npx skills find command. An attacker could publish a malicious skill with instructions in its metadata designed to deceive the agent during the 'Verify Quality' step (Step 4).
Ingestion points: Search results from npx skills find [query].
Boundary markers: None present; the agent processes raw CLI output.
Capability inventory: Subprocess calls via npx, global package installation, and execution of installed skill tools.
Sanitization: The skill suggests manual verification of install counts and source reputation but does not provide technical sanitization or validation of the CLI output before use.

find-skills