The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies on executing various shell commands to perform its curation tasks, including find to locate files, grep and sed to extract metadata, jq to manipulate JSON registry files, and git to commit changes.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from SKILL.md files created in the skills/ directory to update the marketplace registry and documentation.
Ingestion points: The skill reads the name and description fields from the frontmatter of any SKILL.md file found on disk.
Boundary markers: No specific boundary markers or 'ignore' instructions are used during the extraction of metadata.
Capability inventory: The agent has the capability to write to plugin.json and README.md, and can perform git add and git commit operations.
Sanitization: While jq --arg is used to safely update the JSON file, the skill interpolates extracted metadata directly into the README.md documentation, which could be manipulated by malicious content in a skill's metadata.
[EXTERNAL_DOWNLOADS]: The skill instructs the user to run npx skills@latest, which downloads and executes the latest version of the skills package from the npm registry.

skill-curator