LLMSkillHub
mirroredCopied from FerroxLabs/ijfw under MIT. FerroxLabs has not claimed this page, and is not affiliated with LLM SkillHub.

ferroxlabs / ijfw-doctor

Diagnose IJFW integration health per platform. Trigger: 'doctor', 'check setup', /doctor

0
SKILL.md
1.7 KB

IJFW Doctor

Run each check silently, then emit a single positive-framed report. Never say "failed", "error", or "not found" -- reframe every finding as an action or a ready state.

Checks (run in order)

1. MCP Server

Call ijfw_metrics. If it responds: "Memory server active". If unreachable: "Memory server ready to connect -- run claude mcp add ijfw-memory node /path/to/mcp-server/src/index.js".

2. Memory Directory

Check .ijfw/memory/ exists and is writable (test -w .ijfw/memory). If writable: "Memory directory writable". If missing: "Memory directory ready to create -- first memory write will initialize it".

3. Hooks

Check for hook scripts in the platform's hooks directory (e.g. ~/.claude/hooks/ for Claude, .codex/hooks/ for Codex). If scripts exist and are executable: "Hooks registered and live". If missing: "Hooks standing by -- re-run install.sh to register".

4. Platform Integration Depth

Report which features are active based on what's detected:

  • MCP connected = deep integration
  • Hooks present = prompt sharpener active
  • Memory populated = context carry-over active
  • Config present = custom rules loaded

5. Node.js Version

Run node --version. If 18+: "Runtime ready (Node <version>)". If missing or older: "Node.js 18+ needed -- install from https://nodejs.org".

Format

5-line list, one line per check, all positive. End with:

All systems active -- IJFW running at full depth.

Or if any items need action:

IJFW active -- <N> of 5 systems at full depth. Steps above complete the setup.

Under 20 lines total. No negatives.