| name | disk-space-ops |
| description | Check disk usage on a Mac or Linux machine, surface the largest folders and files, and produce conservative cleanup suggestions with risk framing. |
| memory_tags | ["domain:storage","workflow:disk-audit","skill_role:operator","repo_boundary:tools","inputs:filesystem","outputs:cleanup-report","risk:medium"] |
Disk Space Ops
Trigger
Use this skill when the user wants to:
- check disk space or see what is using storage
- find large folders, caches, downloads, backups, or build artifacts
- get practical suggestions for what can be deleted safely versus what needs review first
Do not use this skill for:
- automatic destructive cleanup without explicit user approval
- forensic recovery of deleted files
- low-level partitioning, APFS repair, or disk-health diagnosis
Inputs
- Required: target machine or filesystem context
- Optional: specific path to inspect, workspace/repo path, minimum file size threshold, whether to scan for large files
Workflow
- Start with a non-destructive audit.
Run:
python3 georgeskills/skills/disk-space-ops/scripts/analyze_disk_usage.py- add
--workspace <path> when the user likely has large local dev artifacts in a repo tree
- add
--find-large-files when a file-level scan is worth the extra runtime
- Read the report in this order:
- overall free/used space on the target volume
- largest immediate children of the home directory or target path
- common cleanup hotspots and their current sizes
- large files worth manual review
- Separate recommendations by risk:
- protected: configured MLX Whisper runtimes/model caches or other declared durable toolchains
- safe-ish now: caches, trash, transient build outputs, Xcode derived data
- review first: downloads, local backups, archives, old installers, VM images
- avoid suggesting: system folders, app support data unless the user understands the consequence
- Give the user concrete next actions.
Prefer a short ranked list of the biggest wins, with path, approximate size, why it is large, and whether the deletion is reversible or disruptive.
- Only delete after explicit confirmation.
This skill is for diagnosis and guidance first, not silent cleanup.
Script Surface
Primary entrypoint:
skills/disk-space-ops/scripts/analyze_disk_usage.py
Useful flags:
--target <path>: inspect a different root instead of the current home directory--workspace <path>: scan a repo or workspace for large dev artifacts--top <n>: number of largest entries to print per section--min-file-size-gb <n>: threshold for large-file reporting--find-large-files: include a recursive large-file pass
Outputs
- current free/used disk summary for the target volume
- ranked top folders under the main target path
- a hotspot table of likely cleanup candidates with risk labels
- optional large-file review list
- a short cleanup recommendation set grounded in actual size, not generic advice
Boundaries
- Default to read-only inspection.
- Do not recommend deleting random items under
/System, /private, /usr, or opaque application data without explaining the risk.
- Treat caches and build artifacts as suggestions, not commands, unless the user explicitly asks for deletion.
- Treat
~/.config/georgeskills/mlx-whisper.env as a protection manifest when
present. Never recommend deleting its runtime/cache paths merely because they
are a venv or cache.
- Keep the skill reusable: no hardcoded personal paths, repo names, account ids, or machine-specific assumptions.
