The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection as it processes content from external files.
Ingestion points: The getMarketplaceSkills function in index.ts reads marketplace.json and subsequently reads and parses the content of SKILL.md files from paths specified in the marketplace data.
Boundary markers: None. The skill does not use specific delimiters or instructions to ignore potential commands embedded in the SKILL.md files.
Capability inventory: The skill has permissions to read and write to the local filesystem (specifically installed_skills.json).
Sanitization: None. The skill naively extracts the first header and first non-header line as name and description using regex and string splitting.
Note: This behavior is central to the skill's primary purpose of skill discovery and metadata management.

skill-installer