The Agent Skills Directory

[SAFE]: The skill is a legitimate documentation tool designed to provide a reference menu for an AI agent's local skill library. Its core operations—listing directories and reading metadata from adjacent files—are essential for its functionality and aligned with its stated purpose.
[PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection because it reads and displays untrusted metadata (descriptions and names) from other files in the filesystem.
Ingestion points: Front-matter content of SKILL.md files located in sibling directories (e.g., ~/.claude/skills/).
Boundary markers: Absent; the skill does not wrap the ingested metadata in markers to prevent the agent from interpreting instructions contained within those descriptions.
Capability inventory: The skill is limited to directory enumeration and read-only access to specific metadata fields.
Sanitization: Absent; there is no validation or escaping of the metadata before it is presented to the agent's context.

skills-help