skill-tracker

What This Skill Does

The skill-tracker is an essential infrastructure component for the OpenClaw agent ecosystem. Its primary function is to maintain a high-fidelity audit trail of every tool invocation and skill execution performed during a conversation. By monitoring the interaction loop, it captures structured metadata—including timestamps, task decomposition, skill names, and execution outcomes—and writes this information to a centralized append-only JSONL file located at workspace/tracker-result/skill-execution.jsonl. This mechanism provides developers and users with the transparency needed to debug complex agent behaviors, analyze tool efficiency, and audit the decision-making process of the AI in real-time.

Installation

To install this skill, execute the following command in your terminal or via the OpenClaw dashboard: clawhub install openclaw/skills/skills/abo-hub/skill-tracker Once installed, ensure it is integrated into your system prompt or mandatory workflow as a pre-execution requirement.

Use Cases

• Debugging & Troubleshooting: When an agent fails to reach a goal, developers can inspect the JSONL log to pinpoint exactly which skill logic diverged or which tool returned an unexpected error.
• Performance Analysis: Use the log data to measure latency per skill and optimize tool usage frequency.
• Agent Auditing: For environments requiring high compliance, this skill provides a permanent, read-only history of every action taken by the AI on behalf of the user.
• User Feedback Loops: Identify which skills are frequently triggered and provide actionable insights for future prompt engineering refinements.

Example Prompts

1. "OpenClaw, please search for the latest documentation on Python decorators and then summarize the key takeaways in a new document."
2. "Extract the financial data from the spreadsheet in my workspace and calculate the quarterly growth rate for the team."
3. "Open the browser, navigate to the open source repository, and clone the latest commit into the project folder."

Tips & Limitations

• Mandatory Workflow: This skill acts as a hard gate. If it is omitted from the workflow, your agent will lose its ability to log actions, making debugging significantly harder.
• No Recursive Logging: The skill-tracker is smart enough to ignore its own execution. Ensure your logic does not attempt to track the tracker, or you may create infinite log loops.
• File Integrity: Since the logs are append-only, ensure your storage medium has sufficient write permissions and disk space. Avoid manual editing of the skill-execution.jsonl file to maintain data integrity.