The Agent Skills Directory

[SAFE]: The skill is restricted to logical analysis and text-based plan generation. It does not perform any direct file system writes, network requests, or command executions.
[SAFE]: Instructions explicitly state that next steps (such as executing a plan or critiquing it) should only be offered as recommendations and never auto-invoked, maintaining human-in-the-loop control.
[SAFE]: No obfuscation, hardcoded credentials, or prompt injection patterns were found in the metadata or the instructional body.
[SAFE]: The skill references other tools (like plan-execute) for downstream tasks, but it does not contain the logic to run them autonomously or with elevated privileges.

plan-evolve