What problem does it solve?
RunPod pods often run out of root filesystem or Docker storage space during package installs, image pulls, builds, and long-running experiment sessions, which can break active work and waste paid compute. This skill safely frees up that space without deleting your persistent experiment data stored under /workspace.
Core Features & Use Cases
- Safe, targeted cleanup: Removes only safe-to-delete items like stopped containers, dangling Docker images, package caches, old logs, and build artifacts, with strict rules that never touch /workspace unless you explicitly request it.
- Parallel diagnostic scanning: Spawns independent subagents to audit root filesystem usage, Docker storage, and log/cache sizes in one turn to quickly identify all cleanup candidates.
- Confirmation gates for destructive actions: Requires explicit user approval before pruning Docker volumes, user-created images, or any paths outside the default safe cleanup set.
- Use case: If your RunPod pod is throwing "no space left on device" errors mid-experiment, use this skill to diagnose the source of disk usage and safely reclaim space without losing your training checkpoints or datasets.
Quick Start
Use the sue-runpod-docker-cleanup skill to diagnose and safely clean up RunPod container root disk pressure without deleting your persistent experiment data.
