The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The script scripts/skillhub.sh fetches skill content from the vendor registry at https://skills.droyd.ai/api/skill-content/ and uses an embedded Python script to extract multiple files to /tmp/openclaw-skills/ for agent execution.
[COMMAND_EXECUTION]: The SKILL.md instructions guide the agent to 'follow the loaded skill's instructions' and run bundled scripts from the extracted directory, facilitating the runtime execution of third-party logic.
[PROMPT_INJECTION]: The skill processes untrusted metadata and instructions from a remote API which can influence agent behavior.
Ingestion points: Remote data enters via API endpoints GET /api/search and GET /api/skill-content/.
Boundary markers: The skill does not use specific delimiters or instructions to ignore embedded commands in the fetched data before presenting it to the agent.
Capability inventory: The agent is encouraged to perform file system operations and script execution based on the loaded content.
Sanitization: No sanitization or validation of the downloaded instructions or script content is performed before use.
[EXTERNAL_DOWNLOADS]: The skill downloads data and executable content from the vendor-managed domain https://skills.droyd.ai. While this is an official source for the 'openclaw' vendor, the behavior enables the loading of unvetted third-party content.

skill-search