The Agent Skills Directory

[SAFE]: The skill is a development tool for linting and validating other skill files. It operates locally using standard shell tools and includes explicit security checks against prompt injection techniques (e.g., banning XML tags in metadata).
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted skill files for validation. While this creates an attack surface for indirect prompt injection, the tool itself implements mitigations to detect such patterns in the analyzed files and does not perform any dangerous actions based on the content it reads. Ingestion points: The scripts/validate-skill.sh script reads SKILL.md files from target directories. Boundary markers: None explicitly used in the script output; it uses standard text logging. Capability inventory: Limited to file reading and string processing via standard Linux utilities. Sanitization: Implements a security check that fails if XML angle brackets are found in the target's frontmatter.

skill-linter