The Agent Skills Directory

[SAFE]: The skill serves as a structural alias for the plan skill, implementing a consensus-based deliberation process.
[COMMAND_EXECUTION]: The skill coordinates workflow by invoking other internal platform skills such as plan, ultragoal, and team. No arbitrary or suspicious shell command execution was detected.
[DATA_EXPOSURE]: The skill maintains state by reading and writing to specific workspace directories (.omw/, .workbuddy/). This behavior is restricted to planning artifacts and project metadata, which is consistent with its stated purpose.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data such as user task descriptions and repository contents to generate plans. While it interacts with untrusted input, it does so within a structured multi-agent verification loop (Architect/Critic) which acts as an inherent check on the generated output.

ralplan