The Agent Skills Directory

[SAFE]: The skill instructions focus on standard operational procedures for workflow cancellation. There are no indications of malicious intent or hidden commands.
[COMMAND_EXECUTION]: The skill references specific CLI tools (omh, hermes) for state management and logging. These commands are localized to the OMH environment and are used to query status (omh state status, omh goal board) or clear internal state (omh state clear).
[INDIRECT_PROMPT_INJECTION]: The skill processes user intent to trigger cancellation, creating a surface for indirect injection. However, it incorporates robust safety constraints and boundary markers.
Ingestion points: User-supplied cancellation intent (e.g., "cancel", "stop the workflow").
Boundary markers: Includes a "Do Not Use When" list, a "Completion Checklist", and explicit "Safety rules" that define what the agent should not do.
Capability inventory: Uses omh and hermes CLI commands for state reporting and clearing.
Sanitization: The skill relies on explicit behavioral constraints ("Never end, kill, or signal a host process") rather than technical sanitization of the input string.

omh-cancel