The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill aggregates and reports on data from potentially untrusted internal sources to provide a health dashboard.
Ingestion points: The workflow ingests data from the platform ledger file skills/.curator_ledger.jsonl, lint_warnings generated by the skill_manage tool, and the output of the hermes skills check CLI command.
Boundary markers: Instructions mandate the use of 'not-evidence boundaries' and 'metadata-only wrapper cards' to isolate processed signals from the agent's core instructional context.
Capability inventory: The skill utilizes omh runtime record for telemetry logging and skill_view for reading reference files.
Sanitization: The skill enforces a 'metadata-only' output policy and strictly forbids the agent from claiming or performing automatic modifications to the environment based on the health signals detected.

omh-skill-health