skill-validator
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes metadata from external SKILL.md files. While it uses yaml.safe_load for parsing, the resulting validation report could contain content that influences the agent's behavior if the target file contains prompt injection patterns.\n
- Ingestion points: The script scripts/validate_skill.py reads file content using Path.read_text().\n
- Boundary markers: The script does not utilize specific boundary markers or ignore-instructions delimiters when outputting validation results to the agent.\n
- Capability inventory: The script performs filesystem read operations (os.walk, Path.iterdir, Path.read_text) but lacks network or file-write capabilities.\n
- Sanitization: The script employs yaml.safe_load() for frontmatter parsing and applies strict regex validation to specific fields (name, version, tags).
Audit Metadata