The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill provides a mechanism to fetch and adapt content from external marketplace URLs and GitHub repositories to create new skill definitions. This creates an ingestion surface where instructions from remote sources could influence the agent during the authoring process.
Ingestion points: The skill utilizes an import tool to ingest data from user-provided source URLs (e.g., from skills.sh or specific repositories) into the agent's context.
Boundary markers: The instructions guide the agent to inspect descriptions before importing, but do not provide specific markers or safety delimiters to isolate imported body content from the agent's own instructional hierarchy.
Capability inventory: The skill has access to significant filesystem capabilities through the write, edit, and delete tools.
Sanitization: The skill includes a 'Reminders' section that explicitly warns the user to only include executable scripts in a skill if they have been manually reviewed.
[EXTERNAL_DOWNLOADS]: The skill facilitates downloading content from the skills.sh marketplace and official vendor repositories to support its core authoring functionality.
[COMMAND_EXECUTION]: The instructions recommend using npx skills as a discovery and management method, which involves the execution of external packages via the Node Package Runner.

open-knowledge-write-skill