The Agent Skills Directory

[SAFE]: The skill is designed for administrative cleanup of the AI agent's own work footprint. It includes significant safety constraints to prevent accidental damage or unauthorized access.
[SAFE]: Instructions explicitly forbid the use of recursive deletions on shared directories (e.g., rm -rf on /tmp) and require exact path matching.
[SAFE]: The skill implements a 'proof of ownership' requirement, ensuring the agent only modifies or deletes resources it can verify it created during the current session.
[SAFE]: Process termination is restricted to specific PIDs recorded by the agent, rather than broad name-based matching which could affect other users.
[SAFE]: Sensitive shared resources like package manager caches (npm, pip, etc.) are handled via reporting only, preventing the agent from clearing system-wide caches without explicit user consent.

cleanup