The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill uses npx -y skills-ref validate to download and run the 'skills-ref' utility from the npm registry. This occurs during the final validation step of the skill creation process.
[REMOTE_CODE_EXECUTION]: The use of npx to execute the 'skills-ref' package constitutes remote code execution, as the package is fetched and executed in the local environment at runtime.
[COMMAND_EXECUTION]: The skill executes local Node.js scripts (scripts/validate-skill-metadata.mjs and scripts/validate-skill-links.mjs) to perform structural and metadata checks on the skill being authored.
[INDIRECT_PROMPT_INJECTION]: The skill processes instructions and scripts from other agent skills, creating a surface where malicious content in those files could influence the agent's behavior during the review or improvement process.
Ingestion points: Reads content from SKILL.md, references/, and scripts/ directories of the skill provided for review or creation.
Boundary markers: Absent. The skill does not prescribe specific delimiters or "ignore" instructions for the content being processed.
Capability inventory: Executes shell commands via node and npx to validate inputs.
Sanitization: Absent. The skill relies on specific validation scripts for structural integrity but does not mention sanitizing the natural language content of the skills it processes.

create-skill