The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill uses urllib.request to download markdown definitions from external GitHub repositories. While the default configuration points to trusted organizations, the code provides a mechanism for retrieving content from any repository.
[COMMAND_EXECUTION]: Utilizes npx to execute the skills package directly from a remote registry, which facilitates the discovery of tools within the broader ecosystem but involves running code from an external source.
[PROMPT_INJECTION]: The skill facilitates an attack surface for indirect prompt injection through the processing of untrusted data.
Ingestion points: Content is ingested from external URLs via the download_skill function.
Boundary markers: There are no clear delimiters or instructions that prevent the agent from accidentally following malicious instructions embedded within the downloaded markdown body.
Capability inventory: The skill has the capability to write to the filesystem (Path.write_text), update the local skill registry (skill_registry.yaml), and execute shell commands via npx.
Sanitization: No sanitization or safety filtering is applied to the instruction-bearing markdown content before it is saved and registered for use by the agent.

skill-finder