Skip to content
Back to skills

M3 Health

ASecurity

Health check — package version, installed payload, chatlog DB row count, per-agent hook state.

  • 26 stars
  • 0 votes
  • 0 copies
  • 2 views
  • Added September 19, 2026
ai-agentspythonbashdatabase

Works with

  • cli
  • mcp

Security analysis

A100/100

Scanned September 19, 2026

npx -y skills add skynetcmd/m3-memory --skill m3-health --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of M3 Health?

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

Security grade badge for M3 Health
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/skynetcmd-m3-health/badge)](https://www.skillsdirectory.com/skills/skynetcmd-m3-health)

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: m3-health
description: Health check — package version, installed payload, chatlog DB row count, per-agent hook state.
---
# M3 Health

## When to Use
Use this skill when the user runs a system health check or when you want to diagnose issues with the m3-memory configuration, CLI executables, background services, or databases.

## Instructions
Step 1 — run the doctor command, trying the resolvers below in order. Stop at the first one that returns exit 0; do not run the remaining ones.

```bash
# 1. Plain CLI, if mcp-memory is on PATH:
mcp-memory doctor

# 2. Module form, works whenever the m3_memory package is importable:
python -m m3_memory.cli doctor

# 3. Repo-local venv (developer case, run from the repo root):
.venv/Scripts/python.exe -m m3_memory.cli doctor   # Windows
.venv/bin/python -m m3_memory.cli doctor           # macOS/Linux
```

Step 2 — print the full doctor output verbatim (no paraphrasing).

Step 3 — append exactly ONE short line of interpretation. Examples:
- `all healthy.`
- `chatlog DB never captured — run /m3:install.`
- `Antigravity SessionEnd hook off — run mcp-memory chatlog init --apply-gemini.`

Do not write a paragraph. One line. The user can read the doctor output themselves.

Step 4 — if (and only if) doctor reported a repairable problem (a stale/dead agent
config path, a duplicate bridge, a disabled plugin, or a "run doctor --fix" hint),
OFFER to auto-repair. `doctor --fix` repoints dead config paths, de-duplicates MCP
registrations, and re-syncs agent hooks — it is non-destructive to data. Run the
same resolver that worked above, with `--fix`:

```bash
mcp-memory doctor --fix
# or: python -m m3_memory.cli doctor --fix
```

Then re-run plain `doctor` to confirm the fix took. If doctor was already all
healthy, do NOT run `--fix`.

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.