The Agent Skills Directory

[Indirect Prompt Injection] (LOW): The skill is designed to load and process external metadata from the file system, creating a vulnerability surface where malicious instructions in external files could influence the agent's behavior.
Ingestion points: The loadRegistry function in SKILL.md uses glob and readFile to ingest content from any **/SKILL.md file found in the provided paths.
Boundary markers: The provided code does not show the use of delimiters or specific instructions to the agent to ignore natural language commands embedded within the skill metadata.
Capability inventory: The skill is granted powerful file system tools: Read, Write, Edit, Glob, and Grep.
Sanitization: No sanitization or content validation logic is present in the loading or parsing snippets to filter out executable instructions or malicious prompts within the loaded metadata.

dag-skill-registry