kit-doctorlisted
Install: claude install-skill armenr/modelmux
# /kit-doctor — is this install still live and wired?
A read-only health sweep over a Fieldbook install. It answers one question — *"is everything the kit
promises actually active here, right now?"* — and it catches the silent-inertness failures that don't
throw errors: hooks not wired on a fresh clone, a missing `jq` that no-ops every gate, a Claude Code
version past what the kit was verified against, a manifest that no longer matches disk. It **reports and
suggests**; it changes nothing unless you say yes to a specific offered fix.
**Foreground in the main agent.** Safe to run anytime — it is read-only by default.
## When to invoke
- After a **fresh clone** (hooks are inert until wired — the #1 silent failure).
- After a **Claude Code update** (verify the harness still matches the kit's assumptions).
- When **hooks seem inert** — the safety gate didn't fire, SessionStart said nothing.
- Periodically, or before trusting the install after time away.
## When NOT to invoke
- To reconcile against a newer zip → `/kit-upgrade`. To remove → `concierge/uninstall.md`.
- Mid-destructive-operation — wait for a natural pause (though it's read-only, keep the tree quiet).
## Checks
Run each, collect a `PASS` / `WARN` / `FAIL` with a one-line remedy. Do NOT fix inline — gather all
findings, then offer fixes at the end (§Report).
### 1. Manifest present + parseable
```bash
pwd && git rev-parse --show-toplevel 2>/dev/null
test -f .agent-docs/.kit-manifest.json && \
python3 -c 'im