A comprehensive observability suite providing throughput dashboards, decision auditing, failure tracing, and drift monitoring for AI agents.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install agent-observability
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install agent-observability using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Agent Observability?
Agent Observability is a professional-grade monitoring and logging framework designed specifically for the Openclaw Skills ecosystem. It addresses the 'black box' problem of autonomous agents by providing high-fidelity visibility into their internal logic, costs, and performance quality. By installing this stack, developers gain access to automated dashboards and audit trails that make debugging complex multi-agent workflows significantly more manageable.
The suite is comprised of four specialized tools that work in harmony to track metrics, log high-stakes reasoning, and ensure that agents remain compliant with their original design intentions. Whether you are running agents in a production environment or a local lab, this skill provides the necessary telemetry to scale with confidence.
Agent Observability Use Cases
- Monitoring weekly agent productivity and operational costs.
- Debugging subagent failures by analyzing structured failure traces.
- Auditing high-stakes autonomous decisions to understand the 'why' behind agent actions.
- Ensuring agents follow behavioral guidelines and prevent model drift over time.
- Setting up production-grade monitoring for long-running AI workflows.
How Agent Observability Works
- Installation involves deploying core scripts for dashboards and drift guarding along with library utilities for auditing and tracing.
- Scheduled cron jobs or heartbeats trigger the throughput dashboard and drift-guard scripts to generate weekly performance reports.
- Developers integrate the decision-audit library into agent logic to capture append-only logs of every significant reasoning step.
- If an agent's quality score falls below a specific threshold (e.g., < 7), the failure-tracer automatically captures the context and tool-call sequence.
- All data is persisted into a structured memory directory, allowing for easy querying via standard CLI tools like jq.
Agent Observability Setup
Step 1: Deploy Files
Run the following command to move the observability tools into your workspace:
WORKSPACE="${OPENCLAW_WORKSPACE:-$(pwd)}"
cp references/throughput-dashboard.js "$WORKSPACE/scripts/"
cp references/decision-audit.js "$WORKSPACE/lib/"
cp references/failure-tracer.js "$WORKSPACE/lib/"
cp references/drift-guard-auto.js "$WORKSPACE/scripts/"
Step 2: Configure Automation
Add the reporting tools to your weekly cron or heartbeat script:
node "$WORKSPACE/scripts/throughput-dashboard.js" "$WORKSPACE"
node "$WORKSPACE/scripts/drift-guard-auto.js" "$WORKSPACE"
Agent Observability Data Schema & Taxonomy
The Agent Observability skill organizes data within the memory/ directory using the following structure:
| Path | Data Format | Description |
|---|---|---|
memory/dashboards/ |
Markdown | Weekly productivity and cost snapshots |
memory/drift-reports/ |
Markdown | Compliance scoring against INTENT.md |
memory/decisions-audit.jsonl |
JSONL | Continuous log of agent reasoning and outcomes |
memory/traces/ |
JSON | Individual failure post-mortems and tool sequences |
Agent Observability Advanced Features
- Graceful degradation: Dashboards dynamically adjust content based on available data sources.
- Intent Compliance: Automated scoring for behavioral issues like sycophancy or hallucination hedges.
- Outcome Tracking: Ability to close the loop on decisions by updating log entries with success/failure results.
- Multi-Agent Support: Label-based tracing to isolate failures for specific subagent roles.
- External Integration: Optional session channel logging for monitoring agents across Discord or Slack.
SKILL.md
Loading
Related Openclaw Skills
Kroger Grocery
A specialized skill for managing grocery lists and cart additions across the Kroger family of stores using the kroget CLI.
Logistics Exception Management
A senior-level logistics intelligence skill for managing complex freight exceptions, carrier disputes, and shipping claims across all transport modes.
Open-Meteo Weather
A free, high-performance weather skill that provides current conditions and 7-day forecasts without requiring an API key.
DFMEA - Design Failure Mode and Effects Analysis
A professional quality engineering tool for identifying potential design failures and calculating Risk Priority Numbers in product development.
Deepgram ASR (Nova-3)
A lightning-fast, cost-effective speech-to-text skill leveraging Deepgram Nova-3 for highly accurate multi-language transcription.
MyFood-by Personal Diet Assistant
An AI-powered diet management agent that tracks food inventory and personal dietary restrictions to provide intelligent meal recommendations.
Featured*
