The Agent Skills Directory

[METADATA_POISONING]: The skill's frontmatter contains self-authored security claims, including security_reviewed: true and security_tier: high. These are informational claims by the author and do not constitute an official security certification.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes user-provided JSON annotations and relationship data, representing a potential surface for indirect prompt injection.\n
Ingestion points: Data is ingested through the --annotation, --class, --provided, and --relationships arguments in the Python scripts (schema_checker.py, completeness_checker.py, relationship_checker.py).\n
Boundary markers: The instructions do not define specific delimiters or instructions to ignore embedded commands within the processed data.\n
Capability inventory: The skill uses Bash to run local validation scripts and Read to access ontology metadata files.\n
Sanitization: The scripts implement robust defensive measures, including a 1,000,000-byte cap on input strings, limits on the number of properties processed (max 1000), and validation of class/property names against known ontology terms using difflib and exact match logic.

ontology-validator