Which feedback types are synthesized by the broker?

It synthesizes PR reviews (inspect), issue triage (inspect), autonomous task execution (loop), and automated classification (issue-agent).

How does this skill detect when a human needs to intervene?

The Feedback Broker automatically flags 'escalation_needed' when it detects conditions like three or more polish attempts, multiple 'needs changes' verdicts, or low-confidence automated fixes.

What is the 'handoff gap' in AI development workflows?

The handoff gap refers to the loss of critical context that occurs when one AI agent (like a triage agent) finishes a task and another (like a coding agent) begins, often leading to redundant work or lost requirements.

Can I query feedback for specific pull requests?

Yes, you can use the '--pr' flag followed by the PR number to retrieve a unified view of all inspection feedback and quality dimensions related to that pull request.

Feedback Broker

Name: Feedback Broker
Author: smileynet

bysmileynet

•

生产力与工作流

Synthesizes multi-agent feedback into a unified view to prevent context loss during complex development handoffs.

The Feedback Broker bridges the communication gap between different AI agents within the Line-Cook workflow by aggregating insights from inspection, autonomous loops, and issue classification. It provides a centralized view of historical feedback for issues, pull requests, and tasks, allowing developers and other agents to understand prior attempts, identify escalation needs (such as excessive polish loops), and maintain continuity across the software development lifecycle. By consolidating feedback from sources like Inspect, Loop, and Issue-agent, it ensures that critical context isn't lost when moving between different stages of AI-assisted development.

主要功能

01Multi-dimensional PR review tracking including security, quality, and scope

02Unified context aggregation from Inspect, Loop, and Issue agents

03Structured JSON output for seamless integration with shell scripts and CI/CD

04Automatic escalation detection for stuck tasks and infinite refinement loops

05Queryable history by issue number, task ID, or pull request

066 GitHub stars

使用场景

01Maintaining context when moving from automated issue analysis to manual implementation

02Preventing infinite refinement loops by identifying tasks exceeding three polish attempts

03Aggregating multi-agent feedback into a single report for human code reviewers

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add smileynet/line-cook feedback-broker

For use in Claude.ai and ChatGPT

Download Skill