The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the npx skills command-line tool to perform operations such as searching (find), installing (add), checking, and updating agent extensions.
[REMOTE_CODE_EXECUTION]: The skill provides instructions to install and execute code from remote GitHub repositories or the skills.sh registry using npx skills add <package>. This allows the agent to download and run arbitrary code.
[EXTERNAL_DOWNLOADS]: The skill facilitates the download of external packages and configurations from skills.sh and various GitHub organizations, including vercel-labs and ComposioHQ.
[COMMAND_EXECUTION]: The skill explicitly instructs the agent to use the -y flag (e.g., npx skills add <owner/repo@skill> -g -y) when installing new skills. This flag is used to bypass interactive user confirmation prompts, potentially allowing the agent to install and execute code without human oversight.
[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect injection because it processes untrusted data from an external search registry.
Ingestion points: User-provided queries and external search results returned by the npx skills find command (in SKILL.md).
Boundary markers: None identified; the agent is instructed to present search results and offer installation immediately.
Capability inventory: Command execution and global package installation via the npx skills CLI.
Sanitization: No validation or sanitization is performed on the package names or descriptions retrieved from the external registry before they are presented to the user or passed to the installation command.

find-skills