The Agent Skills Directory

[COMMAND_EXECUTION]: The skill is designed to execute shell commands, specifically npx clawhub@latest install, to facilitate the installation of new agent capabilities. Evidence: Installation and Capability 3 in SKILL.md.
[EXTERNAL_DOWNLOADS]: The skill fetches code packages and binary dependencies like puppeteer and ffmpeg from the ClawHub registry and external repositories. Evidence: Capability 4 in SKILL.md.
[REMOTE_CODE_EXECUTION]: By automating the installation and verification of remote modules, the skill provides a mechanism for remote code execution that is governed by user-confirmation steps. Evidence: Working method and Capability 3 in SKILL.md.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted metadata from external registries to perform semantic matching and prioritization, which is a potential surface for indirect injection. 1. Ingestion points: External registry metadata (Capability 2). 2. Boundary markers: The skill instructs the agent to provide explicit safety warnings for community-sourced content (Boundaries section). 3. Capability inventory: Command execution via npx, file system reads, and network-based catalog search. 4. Sanitization: No explicit sanitization or filtering of registry metadata is described before processing.

find-skills