Broedkrumme Kalibr

What This Skill Does

Broedkrumme Kalibr serves as a critical telemetry and self-improvement layer for your OpenClaw agent. By integrating this skill, users enable the agent to record its decision-making pathways, LLM interaction metadata, and final outcomes. This acts as a feedback loop for the agent, allowing for granular visibility into how specific prompts lead to successful or unsuccessful results. It acts as an "observer" module that monitors operational logs in real-time, providing the data necessary to optimize agent behavior over time through intelligent routing and performance analytics.

Installation

To integrate this telemetry tool into your workspace, use the following terminal command:

clawhub install openclaw/skills/skills/broedkrummen/broedkrumme-kalibr

Once installed, you must configure the skill by adding your unique apiKey and tenantId to your OpenClaw configuration file. Ensure the apiUrl is set to the default provided endpoint unless you are self-hosting your own Kalibr instance. You can toggle features like enableRouting and captureLlmTelemetry to suit your privacy and performance requirements.

Use Cases

• Continuous Agent Training: Use the captured outcome data to identify patterns where the agent frequently fails, allowing for targeted prompt engineering or fine-tuning.
• Performance Auditing: Maintain a persistent log of LLM costs and latency per transaction to ensure your agent stays within budget.
• Outcome Tracking: Monitor the success rate of multi-step automation sequences by logging checkpoints along the execution flow.
• Model Comparison: If intelligent routing is enabled, evaluate which model performs best for specific tasks by analyzing the telemetry generated by each router decision.

Example Prompts

1. "Kalibr, provide a report on the success rate of my last ten research tasks and highlight any recurring failure patterns."
2. "Show me the LLM telemetry logs for the last automation session; I want to see if the model used too many tokens on the data extraction step."
3. "Enable intelligent routing for all subsequent data analysis tasks to optimize for cost and speed."

Tips & Limitations

• Privacy: Since this skill logs LLM interactions, ensure that sensitive or PII data is redacted before being sent to the telemetry endpoint if you are using a third-party hosted instance.
• Storage Costs: Enabling full telemetry capture can result in large log volumes; monitor your storage quotas if you have high-frequency agent usage.
• Optimization: Use the intelligence gained from Kalibr to update your agent's system prompt. The data is only as valuable as the actions you take based on the insights provided.