The Agent Skills Directory

[COMMAND_EXECUTION]: The skill manages planning artifacts and status through vendor CLI commands. Evidence: SKILL.md specifies the use of omh runtime record to track status and omh hermes plan to persist plan data. These tools are used for their designated functions within the oh-my-hermes ecosystem.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted requirements, creating a surface for indirect instructions. Evidence Chain: (1) Ingestion points: user-defined requirements, constraints, and non-goals in SKILL.md. (2) Boundary markers: The skill enforces structured templates requiring stable task IDs and dependencies in SKILL.md. (3) Capability inventory: Subprocess execution of vendor tools and file-write operations to the .hermes/plans/ directory. (4) Sanitization: Mandatory validation against a project-constitution.md document and the requirement for independent verification judges rather than agent self-verification.

omh-plan