Mingjing— AI Agent Health Center

Offline, zero-dependency health monitoring for LLM agents with multi-framework support and a web dashboard for real-time diagnostics and grading.

Pass

Agentic behavior and permission review.

Pattern checks against bundled files.

Multi-engine malware detections and file reputation.

openclaw skills install mngjing

Mingjing — AI Agent Health Center

Paste this into ClawHub publish page. Replace screenshot URLs with actual images.

Zero-dependency, offline-first LLM Agent health monitoring. pip install, zero LLM calls, ~40MB RSS.

Step 1 — Install backend:

pip install mingjing

Step 2 — Enable probe in OpenClaw:

openclaw plugins install mingjing-probe
openclaw config set plugins.entries.mingjing-probe.enabled true
openclaw gateway

Step 3 — Start web panel:

python3 -m src.ming start --daemon
python3 -m src.ming web start --port 18088

Visit http://localhost:18088 to see live data.

Feature	Description
Zero dependencies	Pure Python stdlib, no third-party packages
Offline-first	No outbound traffic, data stored locally
Zero LLM cost	157 rules, pure rule engine, no model calls
7 framework adapters	LangChain / LlamaIndex / CrewAI / OpenHands / Semantic Kernel / AgentScope / Hermes
4-level health grading	Healthy / Sub-healthy / Attention / Critical, per instance
Web dashboard	Live event stream, triage coverage, i18n (EN/中文), dark theme

Layer	Count	Covers
System	20+	Memory, IO, CPU, disk, file descriptors
Probe	8+	Self-check, emit rate, buffer health
Agent	20+	Orchestration, steps, decisions, roles
Model	15+	Tokens, latency, output, frequency
Tool	15+	Duration, errors, output, execution
Network	6+	Connection, DNS, status codes
Security	5+	Injection, sensitive data leaks
Memory	5+	Retrieval, storage, windowing
Plugin	10+	Lifecycle, loading, errors