The Agent Skills Directory

[SAFE]: No malicious instructions, prompt injections, or unauthorized behaviors were detected in the skill's logic or metadata.
[COMMAND_EXECUTION]: The skill defines workflows that execute the vendor's omh CLI tool (e.g., omh runtime efficiency-report, omh runtime record). These commands are used for standard observability tasks such as status recording and report generation from local JSON metadata.
[INDIRECT_PROMPT_INJECTION]: The skill processes external metadata and receipts, presenting a theoretical attack surface for indirect injection.
Ingestion points: Processes run_efficiency_input/v1 files, session_activity_receipt/v1 data, and user-provided context metrics.
Boundary markers: The instructions explicitly mandate the use of "not-evidence boundaries" and require the agent to separate prepared guidance from observed platform/runtime evidence to maintain data integrity.
Capability inventory: The agent can execute omh CLI tools and write reports to the local filesystem using the --write flag.
Sanitization: Input data is restricted to specific JSON schemas (e.g., run_efficiency_input/v1), which mitigates the risk of processing malformed or malicious payloads.

omh-run-efficiency