Proactive Ops Monitor

Proactive operations monitoring for OpenClaw agents. Tracks token utilization, memory layer health, and generates alerts. Provides `/health` dashboard and au...

Security Scan

The skill's files and declared inputs mostly match a local workspace monitoring tool, but the SKILL.md overstates runtime behaviors (continuous monitoring, auto-prepend suggestions, integrations) that are not implemented in the included code and there are a few surprising implementation details to confirm before use.

Purpose & Capability

Name/description (ops monitor, token/utilization, memory health, alerts, suggestions) align with the code's observable behavior: the Python script reads workspace memory files, computes token estimates, generates suggestions, and can write a config. However the SKILL.md claims continuous monitoring and active auto-prepend behavior and third‑party integrations that the script does not implement; those are marketing-level features rather than implemented capabilities.

Instruction Scope

SKILL.md instructs scanning and using workspace memory files (memory/wal.jsonl, memory/working-buffer.md, notes/areas/open-loops.md) which the script indeed reads. Concern: SKILL.md promises continuous background monitoring and automatic prepending of suggestions to agent messages — the provided script is a CLI utility invoked on-demand and contains no hooks to run continuously or to intercept/modify agent outputs. Also, alert logging only occurs if memory/ops-alerts.jsonl already exists (script will not create that file), which contradicts the stated logging behavior.

Install Mechanism

No install spec or external downloads. Code is included directly (scripts/run.py) and uses standard Python libs and local file I/O. No network calls or package installs. Low installation risk.

Credentials

No environment variables, no external credentials requested. The script only accesses files under the workspace (memory directory, config file) and writes its own config and alerts in the workspace; this is consistent with its stated purpose. Note that workspace files may contain sensitive data and the skill will read them.

Persistence & Privilege

Skill does not request always-on or elevated privileges. It writes a config file in the workspace root and may append to memory/ops-alerts.jsonl (but only if that file exists). It does not modify other skills or system-wide settings. Autonomous invocation is allowed by default (not a new risk here) but the code itself is not an autonomous daemon.

What to consider before installing

This skill is not obviously malicious, but treat it as 'works-as-advertised only partially'. Before installing or enabling: 1) Confirm expectations — the included script is a CLI utility (run.py) and will not run continuously or automatically prepend suggestions to agent messages unless your agent runtime invokes it on a schedule or provides hooks. 2) Verify alert behavior — the script only appends alerts if memory/ops-alerts.jsonl already exists (it won't create that file); create a file or adjust the script if you expect automatic logging. 3) Review workspace content — the skill reads memory/* files (which can contain sensitive conversation context or secrets); only run it in workspaces where that access is acceptable. 4) If you need continuous monitoring or integrations (session-wrap-up, agent-oversight), ask the author for a service/daemon implementation or for explicit instructions on how the agent should schedule or invoke the tool. 5) Test in a safe workspace first and consider running under a low-privilege agent account. If the author claims features (auto-prepend, always-on monitoring), request evidence (a background process, webhook endpoints, or integration code) before trusting those behaviors.

Stop being reactive. This skill turns your agent into a proactive partner that warns you before problems hit and suggests next steps.

Capabilities

1. Health Dashboard

/health command or tool("proactive-ops-monitor", "health_dashboard")

Shows:

🛡️  OpenClaw Ops Dashboard
============================
Token Utilization:     62% (74k / 120k)
Memory Layers:
  • WAL entries:       1,234 (12KB)
  • Working buffer:    5.2KB (last 3 turns)
  • Daily logs:        7
Status:               ✅ Healthy

Alerts:
  • None

Suggestions (1):
  1. Session bridge file missing → run /init

2. Token Utilization Tracking

Estimates current context tokens (via char count / 4). Monitors continuously.

Thresholds (configurable):

• Warning: 70%
• Critical: 85%

When threshold crossed:

• Log to memory/ops-alerts.jsonl
• Include suggestion to /wrap_up or compact

3. Alerting

Alerts stored in memory/ops-alerts.jsonl:

{
  "timestamp": "2026-04-01T16:30:00Z",
  "level": "warning",
  "metric": "token_utilization",
  "value": 72,
  "message": "Token usage at 72%. Consider /wrap_up soon."
}

4. Proactive Suggestions

tool("proactive-ops-monitor", "suggest_next", {"limit": 3})

Scans:

• Open loops in memory/wal.jsonl (category=draft, decision)
• Recent conversation gaps
• Unfinished tasks from notes/areas/open-loops.md

Outputs suggested next actions:

[
  {
    "type": "open_loop",
    "content": "Build ollama binary on external machine",
    "context": "mentioned 2 days ago",
    "priority": "high"
  },
  {
    "type": "context",
    "content": "Finish integrating ToolRegistry into core agent"
  }
]

5. Auto-Suggest on Message

When configured, agent will automatically prepend a suggestion to its response if:

• Token utilization > 80%
• There are unresolved open loops
• Working buffer shows recent stalls

Configurable via proactive-ops-config.json.

Configuration

proactive-ops-config.json in workspace root:

{
  "alerts": {
    "token_warning": 70,
    "token_critical": 85,
    "buffer_warning_size_kb": 5000
  },
  "suggestions": {
    "enabled": true,
    "max_per_turn": 1,
    "include_open_loops": true,
    "include_context_gaps": true
  },
  "dashboard": {
    "show_suggestions": true,
    "show_alerts": true
  }
}

Integration

• memory-stack-core: reads WAL for open loops, buffer size for context estimation
• session-wrap-up-premium: monitor suggests wrap-up when utilization high
• agent-oversight: can forward alerts to oversight logs

Usage

Interactive slash commands:

• /health — show dashboard
• /utilization — show token %
• /suggest — show proactive next steps
• /alerts — show recent alerts

Tool calls:

tool("proactive-ops-monitor", "health_dashboard", {"format": "text"})
tool("proactive-ops-monitor", "suggest_next", {"limit": 3})

Pricing

$99/mo includes:

• Real-time monitoring
• Dashboard + alerts
• Proactive suggestions
• Priority support for ops issues

---

Inspired by auto-monitor and proactive-agent-lite from ClawHub.