system-storage-cleanup

Inspect disk usage and propose safe cleanup steps to resolve critical storage pressure.

1.2k|166|Updated Feb 7, 2026
One-click install
npx skills add https://github.com/vellum-ai/vellum-assistant --skill system-storage-cleanup
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: system-storage-cleanup
Source: https://github.com/vellum-ai/vellum-assistant/tree/main/assistant/src/config/bundled-skills/system-storage-cleanup
Command: npx skills add https://github.com/vellum-ai/vellum-assistant --skill system-storage-cleanup

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

When disk space runs critically low, normal assistant work is suspended and you need a safe, guided way to find what is consuming storage and free space without deleting important data.

Core Features & Use Cases

  • Disk Usage Inspection: Runs bounded df -h and du passes on the target volume or workspace to identify the largest directories before drilling deeper.
  • Approval-Gated Cleanup: Presents each proposed deletion with exact path, estimated reclaimable size, and consequence, requiring explicit user approval before removing caches, logs, temp files, or Docker artifacts.
  • Read-Only SQLite Diagnosis: Analyzes data/db/assistant.db growth using read-only PRAGMA and SELECT queries to distinguish table growth from free-page bloat without mutating the database.
  • Use Case: Your assistant locks into storage cleanup mode because the workspace volume is full. The skill inspects /workspace, finds stale Docker artifacts and old logs, and walks you through approving each deletion until the disk-pressure lock clears.

Quick Start

Ask the assistant to inspect what is using disk space in the workspace and propose safe cleanup steps to free storage.

Frequently Asked Questions about system-storage-cleanup

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I find what is using disk space on my workspace?

Run `df -h` on the workspace path to confirm the full volume, then use `du` one directory level at a time sorted by size to locate the largest directories. In Docker mode, focus inspection on the `/workspace` persistent volume.

How to safely free disk space without deleting important files?

Target clearly nonessential items like stale caches, old logs, temporary directories, build artifacts, and Docker cache, and get explicit approval for each exact path before deleting. Never delete credentials, databases, config files, memory data, journals, backups, or user-authored content.

Can I clean up a large SQLite database file manually?

You can inspect it read-only using `PRAGMA page_count`, `freelist_count`, and `dbstat` queries to determine whether growth comes from tables or free-page bloat. Do not run DELETE, VACUUM, or any mutating command; table growth requires product-owned retention work.

When should I use host-level inspection instead of sandbox tools?

Use host-level access only when the sandbox cannot see the volume that is actually full and host inspection is necessary. Explain why host access is needed before using it, and prefer local container-visible inspection first.

What files are protected from deletion during storage cleanup?

Protected items include credentials, security material, workspace database files, config files, migrations, skill and app source, conversation records, memory graph data, journals, reflections, PKB files, and backups. These are only removed if the user explicitly names the path and accepts the consequence.