What does this skill do, and when should you use it?
This is an Agent-driven storage analysis skill: trigger with a natural language phrase, it scans the entire disk (macOS or Windows), finds the biggest space consumers, and classifies each into three tiers — 🟢 safe to auto-clean, 🟡 needs manual judgment, 🔴 caution recommended — with actionable disposal plans. It generates a beautifully formatted, collapsible, interactive HTML report with one-click copy commands and optional web-based deletion (trash or permanent, with confirmation). The entire scan is read-only; nothing is deleted without explicit user action.
Runs a Python script (scan.py) that auto-detects the OS (macOS/Windows) and uses du or os.scandir to compute directory sizes, outputting JSON data. The agent analyzes this data: identifies the top 5 space consumers, investigates mysterious large directories (e.g., UUID-named Containers), and classifies each major item. It then produces an analysis JSON (schema in build_report.py) and either serves it via server.py (local server on 127.0.0.1 with random port and token) to open an interactive HTML report with working delete/trash buttons (using osascript or SHFileOperationW), or generates a static HTML file via build_report.py (no delete buttons).
- When the user says 'My C: drive is full' and needs a quick cleanup plan.
- To find out what is eating 200GB of disk space and get a prioritized list with paths and recommendations.
- To safely clean caches and temporary files on a Mac without risking important data.
- To identify and decide on a mysterious large folder (e.g., a 97GB UUID Container) that appears on disk.
- To produce a shareable disk analysis report as an HTML file.
How do you install this skill?
- The delete functionality relies on a local server whose security isn't independent tested; review whitelist and permissions before use.
- Windows code is untested; exercise caution on first use and backup important data.
- Skill depends on agent compliance with read-only rules; if agent fails to follow, data loss may occur.
- Shell / CLI
Python 3
In Claude Code, Codex, or other Agent Skills-compatible tools, just say:
Install this skill: https://github.com/KKKKhazix/khazix-skills/tree/main/storage-analyzer
The agent will clone it to the appropriate directory. If your agent doesn't support Skills, download the SKILL.md from the storage-analyzer folder and use it as a rules file or paste it into the conversation.
tmp="$(mktemp -d)"
git clone --depth 1 https://github.com/KKKKhazix/khazix-skills.git "$tmp"
mkdir -p ~/.claude/skills
cp -R "$tmp/storage-analyzer" ~/.claude/skills/
rm -rf "$tmp"Generated from the source repository and skill path; it copies only this skill's folder. If the author's install steps above differ, follow those first. To scope it to one project, replace ~/.claude/skills with that project's .claude/skills.
How do you use this skill?
Once installed, send your agent any of these to trigger it:
- Open in Finder
- Move to Trash
- Open in File Manager
After installation, trigger it with a natural language phrase, e.g.:
Check my storage
C drive is full
Clean up my disk
Look at my computer space
storage analysis
The agent will run the scan (may take a few minutes), analyze, and automatically open a browser with the interactive report. In the report, 🟢 items have one-click trash/delete buttons, 🟡 items have 'Open in Finder' and possibly 'Move to Trash' (with confirmation), 🔴 items only have 'Open in File Manager'. The local server runs on 127.0.0.1 with a random port and token for security.
What are this skill's strengths and limitations?
- Fully read-only scan, never deletes anything without explicit user action
- Clear three-tier classification with specific paths and recommendations
- Interactive HTML report with copyable commands and direct action buttons
- Supports macOS (fully tested) and Windows (code ready)
- Pure Python standard library, no external dependencies
- Agent-driven, can identify and explain mysterious directories
- Windows support has not been tested on a real system and may have issues
- Scan can be slow on large disks
- Cleanup commands must be run by the user, or via the web buttons (requires starting the server)
- Requires local Python environment (Windows needs manual install)
How does this skill compare with similar options?
Side by side with related skills; every score comes from the same FSRS standard.
| Skill | FS score | Stars | Last updated | License |
|---|---|---|---|---|
| Storage Analyzer this page | 60 · Recommended | ★ 21k | 10d ago | MIT |
| Cua Driver GUI Automation Skill | 61 · Recommended | ★ 29k | 1d ago | MIT |
| Cua Driver Skill | 58 · Recommended | ★ 29k | 1d ago | MIT |
| Cross-Platform Screenshot Capture ✓ OpenAI · Official | 46 · Use with care | ★ 28k | 3mo ago | — |
| Computer Health Check | 53 · Use with care | ★ 10k | 4d ago | MIT |
How did FollowSkills review this skill?
The skill clearly requires read-only scanning and deletions only through a local server with strict whitelist and confirmation. However, actual enforcement depends on agent compliance and the delete feature isn't independently tested. Thus, full marks not justified; deduct 7 points because of potential bypass risks and reliance on untested mechanisms.
Doc provides clear steps, scripts use stdlib only, but static review can't verify actual operation. macOS claimed tested, Windows not. Happy path plausible, but error handling and edge cases unverified; failures might not be easily diagnosable. Deduct 2 points for insufficient reliability evidence.
Well-defined triggers and boundaries, clear scenarios. Works on macOS/Windows, no overseas service dependency, thus suitable for Chinese users. But there might be unanticipated environments or user expectations; deduct 3 points as full coverage can't be confirmed.
Structure is clear, includes dependencies, platform status, and troubleshooting, but lacks versioning, changelog, and clear update path. Maintenance is personal, not institutional. Deduct 3 points for incomplete governance.
Skill logic sound, value clear, but static review can't measure actual output quality and ease of use. Effectiveness depends on agent analysis, potentially variable. Deduct 8 points due to lack of verification.
No tests or CI in repository for these scripts; README claims testing but no third-party evidence. Thus, only code presence is confirmed, not correctness. Deduct 7 points for insufficient verification.
Open a dimension to read why it scored that way
Evidence confidence：Low — Mostly static review, author material or a limited demo; useful for discovery, not high-risk decisions.
See the full review method →