health

name	health
description	Check Claude Code config health across the team + engine drift
when_to_use	Use when checking engine or config health — runs health.sh (counts vs engine.json, plugin parity, vocabulary sync, installed-agent matrix), reports drift, and reads teammates' statuses from the shared dashboard issue. Triggers on: health, config health, is the engine healthy, drift check, team config status.
argument-hint	[local|all|report]

Check Claude Code config health for all team members.

Arguments: $ARGUMENTS (optional: "local" for local-only check, or "all" for team dashboard)

Steps:

0. Engine consistency (drift check) — bash ~/.claude/scripts/health.sh includes an "engine" section that reads ~/kun/.claude/engine.json (the single source of truth for model + counts) and warns when the docs or repo reality drift from it. It checks: project agent/command/card/rule/MCP counts, stale Opus 4.6/4.7 model refs in docs/ + .claude/CLAUDE.md, and the stale count box in ARCHITECTURE.md. Warnings here mean docs have rotted — update engine.json (if reality changed) or the docs (if they're behind), then re-run. Never fails the gate; warn-only.

1. Run local health check: bash ~/.claude/scripts/health.sh

2. If $ARGUMENTS is "all" or empty:

   • Check GitHub issue databayt/kun with label config-health for latest reports from all team members
   • gh issue list --repo databayt/kun --label config-health --state open --json number,title -q '.[0].number'
   • If issue exists, read comments: gh api repos/databayt/kun/issues/{number}/comments --jq '.[-4:] | .[].body'
   • Summarize: who reported, when, status (healthy/warnings/errors)
   • Flag anyone who hasn't reported in > 7 days

3. If $ARGUMENTS is "report":

   • Run bash ~/.claude/scripts/health.sh --report to post local health to GitHub

4. Present dashboard:

   Member	Status	Last Check	Issues
   Abdout	✅/⚠️/❌	timestamp	details
   Ali	✅/⚠️/❌	timestamp	details
   Samia	✅/⚠️/❌	timestamp	details
   Sedon	✅/⚠️/❌	timestamp	details

5. If any member has errors, suggest: tell [member] to run: cd ~/kun && bash .claude/scripts/setup.sh [role]

Reporting is automatic — the maintain heartbeat supervises every machine:

• setup.sh / setup.ps1 arm a daily scheduler (macOS launchd com.databayt.kun-maintain 09:17 · Windows Scheduled Task kun-maintain · Linux cron)
• Each heartbeat: pull ~/kun → refresh ~/.claude → health check → state to ~/.claude/.kun-maintain.json → post to the dashboard weekly and immediately on RED
• Status on any machine: bash ~/.claude/scripts/maintain.sh --status (Windows: maintain.ps1 -Status)
• A stale heartbeat (>48h) or RED verdict surfaces automatically at session start via the session-maintain-status hook — in every project, not just kun