skill-feedback-collector

What This Skill Does

The skill-feedback-collector is an essential bridge between your OpenClaw agent and your manual oversight. It functions as a sophisticated human-in-the-loop interface that pauses autonomous operations to await user confirmation via a dedicated browser-based UI. Built as an MCP component, it manages a task queue for batch processing, ensuring that you remain in control of complex workflows. When activated, the agent stops and presents a structured summary of its progress, potential risks, or pending decisions, allowing you to guide the development process or verify changes before they become finalized.

Installation

To integrate this skill, navigate to your project directory and run the following in your terminal:

clawhub install openclaw/skills/skills/2019-02-18/skill-feedback-collector

Ensure you have Node.js installed, then execute npm install && npm run build within the skill directory. Configure your MCP settings by pointing the command to build/index.js in the skill's root folder. Finally, access the web interface at http://<server-ip>:18061 to interact with your agent.

Use Cases

• Verification Cycles: Pause after writing code or refactoring modules to inspect results before the agent commits changes.
• Uncertain Parameters: When the agent is unsure about specific UI layout choices, API endpoint naming, or business logic implementation, it will pause for your explicit input.
• Destructive Operation Safeguards: Automatically engage the pause mechanism before dangerous actions like database drops, force pushes, or recursive file deletions.
• Batch Task Management: Feed a sequence of tasks into the internal queue; the agent will process them one-by-one, stopping for your feedback between each transition.

Example Prompts

1. "I am going to start a complex refactor; please use feedback mode so I can approve each module after it is completed."
2. "You have paused because of an uncertain API choice. Let's go with the REST implementation instead of GraphQL for this endpoint."
3. "Switch to free mode now; I trust you to complete the remaining unit tests in the queue without further interruption."

Tips & Limitations

• Proactive Communication: The agent is designed to provide you with a clear menu of options. Always choose the most logical next step to maintain the workflow efficiency.
• Feedback Loop: If you find the agent is pausing too frequently, switch to 'free mode' to regain autonomy. If the process involves high-risk data, keep 'feedback mode' enabled at all times.
• Queue Management: Keep an eye on the task queue size via the UI; if the queue grows too long, it may become difficult to track the context of each pending task.