The Agent Skills Directory

[SAFE]: No malicious patterns, obfuscation, or unauthorized network operations were detected. The skill is an orchestration playbook using specialized tools for project management.
[PROMPT_INJECTION]: The skill utilizes framework-specific role markers (e.g., [omh-role:executor]) to facilitate proper agent behavior via platform hooks. This is a functional requirement and not a safety bypass attempt.
[DATA_EXPOSURE]: File operations are restricted to project-local paths and configuration files within the .omh/ directory and standard git repository structures. No sensitive system or credential paths are accessed.
[INDIRECT_PROMPT_INJECTION]: The skill functions as a dispatcher that aggregates plan data and tool outputs into instructions for subagents.
Ingestion points: Reads plan definitions from .omh/plans/, evidence from omh_gather_evidence, and reports from subagent executions.
Boundary markers: Uses the structured markdown templates in the references/ directory to clearly separate instructions from processed data.
Capability inventory: Utilizes delegate_task for spawning subagents and omh_gather_evidence for running project-specific test suites.
Sanitization: Relies on a verification subagent to validate the correctness of outputs but lacks explicit character escaping for interpolated plan descriptions.

omh-ralph-driver