reclaim
Installation
SKILL.md
Reclaim
Run a manual, read-only storage audit. Find what occupies space, why it exists, whether it is still required, and what prevents repeated growth. Return a report. Do not equate a large folder or an old timestamp with waste.
Scope and boundaries
- Use the user's requested machine or project scope. For a whole-machine request, inspect local volumes, common user folders, developer projects, application data, and background services. For a project request, inspect its dependencies and output plus relevant shared stores; do not expand into unrelated personal files.
- Do not delete, move to Trash, uninstall, prune, upgrade, edit configuration, stop or start services, empty Trash, or install automation. A later cleanup request is a separate action, not part of invoking this audit.
- Use installed inspection tools. Do not install a scanner, run builds, download dependencies, or execute project scripts just to measure storage.
- Keep inspection local. Verify the active container endpoint before any container query. Do not contact a remote daemon or inspect production settings unless the user requests that scope.
- List blocked paths and stopped services as coverage gaps. Do not bypass permissions or start a stopped runtime to complete the report.
- Read metadata and manifests before file contents. Do not print credentials, private keys, browser records, database rows, or memory-dump contents. Avoid recursive searches through secrets and private documents.
Audit sequence
1. Establish a baseline
Identify the operating system, local volumes, free space, and requested roots. Record the measurement time and units. Use a shallow disk-usage scan, then drill into the largest relevant children. Include hidden directories and ignored build output where appropriate; a Git-tracked file list is not a disk inventory.