The Agent Skills Directory

[Prompt Injection] (SAFE): No direct attempts to override system instructions or bypass safety filters were detected in the prompt or code snippets.
[Indirect Prompt Injection] (LOW): The skill identifies a mechanism for ingesting untrusted data from the file system, which could be exploited to influence agent behavior.
Ingestion points: The loadRegistry function uses glob to find and readFile to ingest content from multiple SKILL.md files.
Boundary markers: The provided logic does not include delimiters or specific instructions to the agent to ignore potential commands embedded within the parsed metadata.
Capability inventory: The skill has access to sensitive tools including Read, Write, Edit, Glob, and Grep.
Sanitization: No sanitization or validation logic is shown for the parseSkillFile process to prevent instruction injection from the cataloged files.

dag-skill-registry