One-click install
npx skills add https://github.com/redhat-community-ai-tools/UnifAI --skill pipeline-debugger-redhat-community-ai-tools
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: pipeline-debugger
Source: https://github.com/redhat-community-ai-tools/UnifAI/tree/main/.cursor/skills/pipeline-debugger
Command: npx skills add https://github.com/redhat-community-ai-tools/UnifAI --skill pipeline-debugger-redhat-community-ai-tools

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

It helps teams debug multi-agent pipeline errors, behavioral mismatches, and performance regressions by turning unclear symptoms into reproducible evidence, isolated root causes, and targeted, architecture-respecting changes.

Core Features & Use Cases

  • Structured, evidence-first debugging: Follows a strict 6-step workflow to gather evidence, reproduce issues, isolate the fault, diagnose the root cause, propose a fix, and verify outcomes.
  • Pipeline-aware investigation: Uses prior pipeline phase outputs and revision loop feedback to avoid repeating already-attempted guesses.
  • Hexagonal architecture compliance: Ensures that fixes correct architectural violations rather than working around them.
  • Use Case: When a pipeline gets stuck after exhausting revision loops, it systematically narrows the failure to a specific file/function and architectural layer before proposing the minimal fix.

Quick Start

Use the pipeline-debugger skill when you have an error message, expected vs actual behavior, or performance observation and you need a reproducible, code-referenced root-cause analysis followed by a targeted fix.

Frequently Asked Questions about pipeline-debugger

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I debug multi-agent pipeline failures using logs and stack traces?

Debug multi-agent pipeline failures by following a strict evidence-first workflow that gathers logs and stack traces, reproduces the issue, isolates the root cause to specific architectural layers, and proposes a minimal verified fix.

What is the best way to isolate root cause when a multi-agent pipeline gets stuck?

The best way to isolate root cause is using pipeline-aware investigation that analyzes prior phase outputs and revision loop feedback to narrow failures down to a specific file, function, and architectural layer before proposing targeted fixes.

How do I fix behavioral mismatches without violating hexagonal architecture?

Fix behavioral mismatches by diagnosing the root cause and proposing a minimal fix that respects hexagonal architecture, ensuring corrections address architectural violations directly rather than working around them, followed by test verification.

Can I use evidence-first debugging for performance profiling and performance regressions?

Yes, evidence-first debugging applies to performance profiling and performance regressions by turning unclear performance observations into reproducible evidence, isolated root causes, and targeted architecture-respecting changes.

Why does my pipeline keep failing after exhausting revision loops?

Pipelines fail after exhausting revision loops because prior guesses were not systematically narrowed. A disciplined debugger uses prior phase outputs to isolate the fault to a specific location and architectural layer before applying a minimal fix.