The Agent Skills Directory

[SAFE]: The skill is a legitimate workflow automation tool. It defines a strict operational sequence and uses project-local storage for context and state tracking.- [COMMAND_EXECUTION]: The skill invokes other workflow phases ($ralplan, $ralph, $code-review) and utilizes the omx command-line utility for state persistence. These operations are restricted to the vendor's own ecosystem and expected developer tasks.- [PROMPT_INJECTION]: The skill's autonomous nature presents an indirect prompt injection surface through the ingestion of task descriptions or external issue content. However, the multi-stage architecture with a separate code-review gate mitigates this risk by ensuring implementation results are verified before completion.
Ingestion points: User requests, GitHub issues, and .omx/context/ files (SKILL.md).
Boundary markers: None explicitly identified.
Capability inventory: Local file writes, state management via omx CLI, and implementation execution via $ralph (SKILL.md).
Sanitization: None explicitly mentioned.

autopilot