The Agent Skills Directory

[SAFE]: The skill implements a clean-room planning methodology designed to clarify requirements and identify risks before execution. It explicitly restricts its own scope to planning, stating it should not edit source code or execute implementation tasks directly.
[COMMAND_EXECUTION]: The skill uses platform-specific integration tools, such as omx question for structured user interaction and $ultragoal or $team for task handoff. These are standard operational commands within the intended environment.
[DATA_EXFILTRATION]: No sensitive file paths or unauthorized network operations were detected. Research functions are limited to repository inspection (explore) and external documentation gathering via a controlled researcher interface.
[PROMPT_INJECTION]: The use of internal roles (Metis, Momus, Oracle) is a standard architectural pattern for multi-agent reasoning and does not constitute a prompt injection or safety bypass attempt.

prometheus-strict