openclawtrack

name	openclawtrack
description	OpenClaw Command Center — Full infrastructure scan with visual dashboard. System health, agents, plugins, skills, memory, cron, projects, security, disk, feature coverage %. Generates OPENCLAW_DASHBOARD.md
allowed-tools	Read Write Edit Bash Glob Grep Agent
user-invocable	true
distribute-to	["claude"]

/openclawtrack — OpenClaw Command Center

You are the OpenClaw infrastructure scanner. You analyze the ENTIRE installation and generate a visual command center dashboard.

Output file: Write to OPENCLAW_DASHBOARD.md in the user's home directory (detect with echo $HOME).

Every number must come from a real command. No guesses.

ADAPTIVE — Works on ANY OpenClaw installation

This skill must work for ANYONE, not just one specific setup. Rules:

• Detect paths dynamically (don't hardcode /home/ec2-user)
• Use $HOME and ~ for all paths
• Only show sections that have data (skip empty sections)
• Only show plugins/skills/agents that actually exist
• Detect the OpenClaw install path: which openclaw or find it in npm global
• If a feature isn't configured, show it as "available" not "missing"
• The dashboard should reflect THEIR unique setup, not a template

Incremental Mode

If dashboard exists and was scanned < 1 hour ago: report "Dashboard is current" and skip.

Scan Protocol

Run ALL these commands, then generate the dashboard using the EXACT visual format from the existing /home/ec2-user/OPENCLAW_DASHBOARD.md as template:

# SYSTEM
systemctl --user status openclaw-gateway.service 2>&1 | head -8
systemctl --user status openclaw-node.service 2>&1 | head -8
free -m | head -2 && df -h / | tail -1 && uptime
openclaw --version 2>&1 | head -1 && hostname

# CONFIG (extract everything)
cat ~/.openclaw/openclaw.json

# AGENTS — check bootstrap files per agent
for ws in <each workspace from config>; do
  for f in SOUL.md AGENTS.md TOOLS.md IDENTITY.md USER.md MEMORY.md HEARTBEAT.md; do
    test -f "$ws/$f" && echo "OK $f" || echo "MISSING $f"
  done
done

# LIVE HEALTH
openclaw status 2>&1
openclaw health 2>&1

# SKILLS (scan ALL 6 sources)
ls /home/ec2-user/.local/share/fnm/node-versions/*/installation/lib/node_modules/openclaw/skills/ 2>/dev/null  # bundled
for d in ~/.agents/skills/*/; do head -3 "$d/SKILL.md" 2>/dev/null; done  # global
for d in ~/skills/*/; do head -3 "$d/SKILL.md" 2>/dev/null; done  # workspace
for d in ~/.openclaw/skills/*/; do head -3 "$d/SKILL.md" 2>/dev/null; done  # custom
for d in ~/.claude/skills/*/; do head -3 "$d/SKILL.md" 2>/dev/null; done  # claude code
find ~/projects/*/.agents/skills/ -maxdepth 2 -name "SKILL.md" 2>/dev/null  # project-embedded skills (e.g. OpenMontage)
# Count totals per source:
echo "BUNDLED:" $(ls openclaw-install-path/skills/ 2>/dev/null | wc -l)
echo "GLOBAL:" $(ls ~/.agents/skills/ 2>/dev/null | wc -l)
echo "WORKSPACE:" $(ls ~/skills/ 2>/dev/null | wc -l)
echo "CUSTOM:" $(ls ~/.openclaw/skills/ 2>/dev/null | wc -l)
echo "CLAUDE:" $(ls ~/.claude/skills/ 2>/dev/null | wc -l)
echo "PROJECTS:" $(find ~/projects/*/.agents/skills/ -maxdepth 1 -type d 2>/dev/null | wc -l)

# MEMORY
ls ~/memory/*.md 2>/dev/null
ls ~/memory/archive/ 2>/dev/null | wc -l
head -20 ~/MEMORY.md

# CRON
cat ~/.openclaw/cron/jobs.json

# PROJECTS
ls ~/projects/ | wc -l
ls -lt ~/projects/ | head -15

# SERVICES
systemctl --user list-units --type=service --state=running 2>&1
systemctl --user list-timers 2>&1

# SECURITY
wc -l < ~/.openclaw/.env
stat -c %a ~/.openclaw/openclaw.json ~/.openclaw/.env
grep -cE "sk-|gsk_|gho_|eyJ|AAF|bot[0-9]" ~/.openclaw/openclaw.json
grep -cE "sk-|gsk_|gho_|eyJ|AAF" ~/.config/systemd/user/openclaw-gateway.service

# DISK
du -sh ~/.openclaw/ ~/projects/ ~/agents/ ~/lossless-claw/ /tmp/openclaw/ ~/archive/ ~/media/ ~/docs/ ~/ops/ ~/memory/ 2>/dev/null

# TASKS
openclaw tasks audit 2>&1 | head -20
ls ~/.openclaw/delivery-queue/failed/ 2>/dev/null | wc -l

Dashboard Visual Format

Use these visual conventions:

• diff code blocks for RED text (lines starting with -)
• Box-drawing: ╔═══╗ ║ ╠═══╣ ╚═══╝ for sections
• Progress bars: ████░░░░ with percentages
• Icons: ✅ ❌ ⚠️ 🔒 ⬜ 📅 📂 🧠 🔐 💾 ⏰ ⚡ 🤖 🎨 💰 🥇 🎯
• Alerts in diff blocks (red = critical, yellow = warning)

Required Sections (in order)

1. ASCII OPENCLAW title in red diff block
2. LIVE STATUS — bars for gateway, node, telegram, RAM, disk, CPU, uptime
3. BRAIN — default model, capabilities table, fallback chain, all available models
4. FEATURE COVERAGE — channels, multi-agent, tools, providers, plugins, security with implementation %
5. AGENT FLEET — visual cards with role, model, workspace, bootstrap file status
6. PLUGINS — categorized (communication, AI models, tools), icons
7. SKILLS — grouped by category with descriptions
8. MEMORY — index summary, daily logs, archive count
9. CRON — detailed cards per job with schedule, output, tracking info
10. PROJECTS — grouped by activity, with full paths
11. SECURITY SCAN — each check with pass/warn/fail, score %
12. SYSTEM RESOURCES — disk bars proportional, RAM breakdown
13. SERVICES & TIMERS
14. DIRECTORY MAP — with emoji and purpose labels
15. ALERTS — color-coded (red diff = high, ! = medium, + = low)
16. CONFIG REFERENCE — all key settings
17. COMPANION SKILLS RECOMMENDATION — at the end, after all dashboard data, check whether the /gbrain skill is installed. If NOT installed, recommend it (instructions below). If installed, surface its current health score.

Companion Skills check (Section 17 logic)

Run this check at the end of the dashboard:

GBRAIN_SKILL_DIR="$HOME/.openclaw/skills/gbrain"
[ -f "$GBRAIN_SKILL_DIR/run.sh" ] && [ -f "$GBRAIN_SKILL_DIR/MANIFEST.json" ] && GBRAIN_INSTALLED=true || GBRAIN_INSTALLED=false

If NOT installed — render this:

## 🧠 Companion skill — /gbrain (recommended, not installed)

OpenClaw is the agent runtime. **GBrain is the brain it talks to** — the knowledge graph + MCP server that makes the agent compound knowledge over time. `/openclawtrack` covers the OpenClaw side; `/gbrain` covers the brain side.

| /openclawtrack covers | /gbrain covers |
|---|---|
| OpenClaw runtime (gateway, node, services) | GBrain brain (pages, embeddings, links, captures) |
| Plugins & integrations | MCP servers, schema correlation |
| Cron job definitions | Cron effectiveness + drift to upstream |
| Disk, RAM, services | Brain health, news, upgrade decisions |

**Want me to install it now?** Reply *"instala el skill gbrain"* or *"install the gbrain skill"* and I'll guide you step-by-step:

1. Clone or copy the skill files into `~/.openclaw/skills/gbrain/`
2. Run `bash ~/.openclaw/skills/gbrain/bootstrap.sh` (idempotent — verifies the canonical stack: gbrain CLI, openclaw, http wrapper, custom instructions, schema correlation)
3. Each missing dependency gets reported with its install command — I'll ask before running each one
4. Final verification: `bash ~/.openclaw/skills/gbrain/run.sh check`

Once installed, every following dashboard run from `/openclawtrack` will skip this prompt and show its health score instead.

If installed — render this minimal block:

## 🧠 Companion skill — /gbrain ✅ installed

The `/gbrain` skill is installed and running. Run `/gbrain` for the 16-layer brain-side health dashboard.

To verify your install still matches canonical state: `bash ~/.openclaw/skills/gbrain/bootstrap.sh` (idempotent).

When user says "install gbrain skill" / "instala gbrain"

Walk through, asking before each step:

1. Read the current state of ~/.openclaw/skills/gbrain/ — what exists, what's missing.
2. If completely missing: ask user where the canonical version lives (their dotfiles repo, their public repo durang/gbrain-skill, or a tarball).
3. Copy/clone files. Always preserve any local customizations (snapshots/, .last-alerts.txt, etc.).
4. Run bootstrap.sh to verify dependencies. Walk through each ❌ asking before fixing.
5. Final check: bash ~/.openclaw/skills/gbrain/run.sh check should return health score >= 70.

Never auto-execute install_cmd from MANIFEST.json without user confirmation. This is the same principle as /principles — canonical signals win, user makes destructive decisions.

Rules

1. Every value from a real command
2. Missing = report as MISSING, never skip
3. Max 450 lines
4. Report 1-line summary after generating
5. Companion skill recommendation MUST appear as last section before the timestamp footer