skills-doctor
Skill11 starsChanged 14 days agoMITView on GitHub
Use when diagnosing the tools, operator identities, API scopes, resource roles, runtime dependencies, or approvals needed to execute skills, especially when handing a workflow to a teammate.
- Reads credentials
4 other projects use a version of this.
Adopted by
Most recently taken up by
- wangyuchen5110/kimi-skills-doctortheir own version, first seen Nov 8, 2026
- thititongumpun/skillstheir own version, first seen Nov 2, 2026
- milad-ahmd/agent-skillstheir own version, first seen Nov 1, 2026
- leacent/skills-doctortheir own version, first seen May 16, 2026
Other versions4 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 4
- milad-ahmd/agent-skills★ 5same name
- leacent/skills-doctor★ 1same name
- thititongumpun/skills★ 1same name
- wangyuchen5110/kimi-skills-doctor★ 1same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository19 other files it gives its agents.
- AGENTS.mdAGENTS.md
- obsidian-skills/AGENTS.mdAGENTS.md
- add-just-doctorSkill
- agent-autoresearchSkill
- agent-harness-repairSkill
- change-evidenceSkill
- ci-speedSkill
- verifySkill
- deck-style-gallerySkill
- doneSkill
- longform-fictionSkill
- loop-kitSkill
- loop-nightly-reviewSkill
- loop-triage-githubSkill
- loop-watch-socialSkill
- orch-claude-pi-codexSkill
- tdd-pstackSkill
- teach-pstackSkill
- windows-leftover-cleanupSkill
The file
We split this skill into the rules it gives an agent
4 rules. 0 are common, 3 are distinctive, and 1 is very specific to this project.
0
Common
Most of the 3 edited versions we read keep these.
3
Distinctive
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 37
Read the selected skill, its relevant references, and required external skills.
- Distinctive: line 65
Read the target repository's live-safety instructions first.
- Distinctive: line 89
Keep current-operator findings separate from the colleague's requirements.
- Distinctive: line 95
Use a private, ignored location for saved reports.
“Common” is measured: kept by at least 6 in 10 of the 3 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Skills Doctor
Diagnose prerequisites, not business outcomes. Run under the intended operator's own account. An administrator's successful run does not establish a colleague's readiness.
1. Inventory The Target
Locate this installed skill's bundled runner and select the checkout containing the skills to diagnose:
python3 <installed-skill>/scripts/doctor.py --repo-root <checkout> --all
python3 <installed-skill>/scripts/doctor.py --repo-root <checkout> --skill <name>The runner uses Python 3.10+ and the standard library. It defaults to the caller's
working directory, not its own installation. Discovery uses the first existing
.claude/skills, .agents/skills, or .codex/skills directory, falling back to
the checkout root. Set repeatable --skills-dir paths to override that scope;
use --skills-dir . for a whole skill-library checkout.
Without a catalog, each skill remains unverified. Inline command mentions
are candidates, not proof of required tools or permissions. Offline discovery
does not run those commands, load .env, inspect credentials, or contact services.
Done: name the selected skills and distinguish observed tool hints from reviewed requirements. Report unresolved dependencies rather than guessing.
2. Review Requirements
Read the selected skill, its relevant references, and required external skills. For each execution branch, identify:
- Local executables and interpreter/package versions.
- Operator identity, API scopes, resource roles, and network reachability.
- External skill packages, agent Web/MCP tools, browser sessions and remote runtimes.
- The separate human approval needed for actual business operations.
Record the requirements in a target-local catalog using the catalog contract and example. Keep private hosts, employee data, credentials, device paths and live reports in the consuming repository, not in this reusable skill. A narrower branch must reflect the task: onboarding marketing employees need not require host or AWS administration; already-provisioned VPN access need not require VPN administrator rights.
python3 <installed-skill>/scripts/doctor.py --repo-root <checkout> \
--catalog .skills-doctor.json --validate-catalog
python3 <installed-skill>/scripts/doctor.py --repo-root <checkout> \
--catalog .skills-doctor.json --skill <name> --profile <branch>Done: every selected capability has evidence sources, a granting owner, and a repair action. Employee access being provisioned is distinct from the operator permissions needed to provision it.
3. Probe Only The Authorized Scope
Read the target repository's live-safety instructions first. Live checks require an explicitly selected skill, a reviewed catalog, and selected probe families:
python3 <installed-skill>/scripts/doctor.py --repo-root <checkout> \
--catalog .skills-doctor.json --skill <name> --profile <branch> \
--live github --github-org <organization>For exact flags, evidence and limitations, read the diagnostic contract. Available probes cover Lark user scopes, GitLab administrator access/token scope, GitHub organization ownership/token scope, AWS identity, and specific Kubernetes RBAC actions. AWS identity alone does not prove IAM or billing permissions.
Manual capabilities stay unverified. The runner never attempts SSH/sudo/Rails fallbacks, grants, login flows, messages, invitations or deployments.
Done: each capability is verified for its stated evidence, known blocked, or explicitly unverified. Do not execute the business workflow merely to test access.
4. Hand Off
Report the tools, roles, remaining blockers, owners and minimum next actions. Keep current-operator findings separate from the colleague's requirements. Tool installation, logins, credential consent, privilege grants and business operations require their own authorization. Use colleague-owned credentials; administrator token/key/session copying is not a repair.
The default output is Markdown; --format json emits the same diagnostic data.
Use a private, ignored location for saved reports. Exit 2 means known blockers
or invalid input; exit 3 means unresolved checks; exit 0 is not business approval.
Maintenance
The reusable skill ships the runner, fixed probes, and synthetic tests. Consumer catalogs remain consumer-owned; the public example contains no organization inventory. After changing the runner, run:
python3 -m unittest discover -s <installed-skill>/tests -p 'test_*.py'Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

