Skip to content
Back to skills

Metis Doctor

ASecurity

metis doctor, health check, diagnose, what's broken, troubleshoot, system status, self-test, sanity check, metis-doctor

  • 3 stars
  • 0 votes
  • 0 copies
  • 0 views
  • Added September 23, 2026
ai-agentspythongobashsqlgitapidatabase

Works with

  • claude desktop
  • api
  • mcp

Security analysis

A100/100

Scanned September 23, 2026

npx -y skills add SVerITG/Metis --skill metis-doctor --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Metis Doctor?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Metis Doctor
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/sveritg-metis-doctor-metis/badge)](https://www.skillsdirectory.com/skills/sveritg-metis-doctor-metis)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: Metis Doctor
description: "metis doctor, health check, diagnose, what's broken, troubleshoot, system status, self-test, sanity check, metis-doctor"
model: claude-haiku-4-5-20251001
effort: normal
complexity: quick
---

## Purpose

Run a one-screen health check on Metis. Reports on Python, the SQLite database, the Anthropic API key, your `user-config.yaml`, agent and skill folders, folder-rename hygiene, MCP imports, and `.env` git safety.

Use whenever:
- Something feels broken and you want a single command to triage.
- You just `git pull`-ed and want to confirm nothing regressed.
- You're about to publish or share the repo and want to catch hygiene issues.

## What to do when invoked

Call the MCP tool `metis_doctor` directly. It returns a Metis-branded status block (the same `Metis · Research Cortex` look as the dashboard, with ✓ / ⚠ / ✗ status glyphs). **Render it as-is — do not paraphrase.** Then, for every ⚠ or ✗ row, add a short plain-language paragraph telling the user what it means and the exact next step — in researcher language, never developer jargon.

The check covers: Python, the database, the Anthropic API key, the **embedding/RAG engine** (fastembed + sqlite-vec), the **knowledge layer** (are docs indexed?), agents/skills, the **dashboard on :8080**, the **Claude Desktop link** (is metis-rc registered?), folder hygiene, MCP imports, and `.env` git safety.

## Output format

```
════════════════════════════════════════════════════
  Metis · Research Cortex — Health Check
════════════════════════════════════════════════════
  Status: ✓ HEALTHY  |  ⚠ NEEDS ATTENTION  |  ✗ PROBLEMS FOUND
  <one-line summary>

  ✓  Python 3.10+              running 3.12.3
  ✓  Embedding / RAG engine    fastembed + sqlite-vec installed
  ⚠  Anthropic API key         not found in env or .env
  ⚠  Claude Desktop link       found but metis-rc not registered — re-run setup-mcp.sh
  ...
────────────────────────────────────────────────────
  Next step → <most severe open item, in plain language>
```

Translate each open item into a plain fix, e.g.:
- *Embedding / RAG engine missing* → "Semantic search is off. Reinstall Metis with the embedding extra: `bash tools/reinstall-mcp.sh`."
- *Anthropic API key* → "Run `/metis_config` (or `/metis-customize`) to paste your key."
- *Claude Desktop link* → "Re-run `bash system/mcp-server/setup-mcp.sh`, then quit and reopen Claude Desktop."
- *Dashboard not running* → "Start it from the Metis desktop icon, or run `bash system/app-py/run.sh`."

## Edge cases

- Doctor itself fails to import: tell the user the MCP server may not be installed yet — direct to `bash setup-mcp.sh`.
- All checks pass but user reports something is broken: ask them what surface is misbehaving (which tab, which command) — Doctor is broad, not deep.

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.