Skill Doctor
Skill Manager
Turn a scattered collection of local AI Skills into a clear, evidence-backed system you can actually manage.
What it does
- Discovers Skills across authorized Codex, Claude Code, and project directories.
- Separates physical presence, host-confirmed discoverability, inferred discoverability, and runtime-only entries.
- Detects exact duplicates, same-name variants, related Skill suites, placement issues, and maintenance risks.
- Compares the current scan with previous snapshots to show what was added, removed, moved, or changed.
- Produces an offline interactive HTML management map together with a plain-text summary.
- Builds reversible action plans; no real file change happens without an exact preview and your explicit approval.
Typical use cases
- Understand which installed Skills are actually available to your AI client.
- Find duplicate, outdated, misplaced, or conflicting Skills.
- Separate user-wide Skills from project-specific Skills.
- Keep related Skill suites together instead of managing them as disconnected files.
- Review how your local Skills setup changed over time.
How it works
- Introduces an optional management framework and asks for permission before scanning.
- Performs a read-only scan limited to the directories you authorize.
- Builds an evidence-backed inventory and diagnoses management issues.
- Generates an offline interactive report plus a text summary.
- Shows proposed actions one by one. Only actions you explicitly approve can be applied.
- Re-scans after approved changes and reports the result.
Example
Input
Audit the AI Skills in this project. Show me what is active, duplicated, misplaced, or changed, and generate an offline management map.
Output
- A text summary with physical, confirmed, inferred, and independent Skill counts.
- An offline interactive HTML report with search, filters, scope views, related-suite views, findings, and change history.
- A list of recommended actions with evidence and confidence levels.
- No file modifications unless you separately approve a specific action plan.
Privacy and safety
- Runs locally on the buyer's computer.
- Does not upload local Skill inventories, paths, or reports to the publisher.
- Scanning is read-only and limited to explicitly authorized directories.
- External links, referenced files, host configuration, and Git evidence require separate authorization when they fall outside the approved scope.
- The product never deletes user files. Removal actions use recoverable archival or unlink only the approved discovery entry.
Requirements and compatibility
- Python 3 using only the standard library.
- No cloud account, API key, external service, CDN, or third-party runtime dependency is required.
- macOS is fully validated.
- Linux and Windows support is experimental and has not been fully validated on real devices.
FAQ
Is this a security scanner?
No. Its primary purpose is Skill discovery, scope visibility, duplicate and variant management, placement, maintenance, and change tracking. Basic safety warnings are secondary.
Will it change my files during a scan?
No. Scans are read-only. Every real change requires a separate preview and explicit approval.
Does it send my local data anywhere?
No. The inventory and reports remain on your device.
Can I modify the downloaded files?
Yes, for your own personal use or your own authorized business use. Resale, redistribution, sharing, republication, sublicensing, and use for AI model training are prohibited by the included license.
