finding-code-bugs

Identify and reproduce software defects using tests, logs, and diffs.

Updated Apr 25, 2026
One-click install
npx skills add https://github.com/Sailfishc/guanfu --skill finding-code-bugs
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: finding-code-bugs
Source: https://github.com/Sailfishc/guanfu/tree/main/skills/finding-code-bugs-skill
Command: npx skills add https://github.com/Sailfishc/guanfu --skill finding-code-bugs

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Investigate and resolve software defects by constructing concrete problems, outlining reproduction steps, collecting evidence (logs, diffs, tests), and guiding a principled fix with validation.

Core Features & Use Cases

  • Problem Card creation and guided workflow for bug hunts
  • Reproduction-first approach using concrete evidence
  • Root-cause analysis, targeted fixes, and regression verification
  • Applicable to crashes, failing tests, flaky behavior, and suspicious diffs

Quick Start

Provide a clear bug report with a reproducible scenario and follow the guided workflow to isolate and verify the root cause.

Frequently Asked Questions about finding-code-bugs

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

FAQPage Schema
What is evidence-driven debugging and how does it help find code bugs?

Evidence-driven debugging identifies and reproduces software defects by constructing concrete problem statements using logs, diffs, and tests. It bounds the scope of failures and guides fixes with minimal, verifiable changes validated by regression checks.

How do I reproduce a flaky behavior or failing test using a structured workflow?

To reproduce failing tests or flaky behavior, create a problem card with a reproducible scenario, collect concrete evidence like logs and diffs, and coordinate a reproduction plan to confirm failures before isolating the root cause.

How do I isolate a root cause from suspicious code diffs and logs?

Isolate a root cause by analyzing suspicious diffs and logs to construct concrete problem statements. Use this evidence to bound the failure scope, confirm the defect, and guide a targeted, minimal fix.

Can I use this reproduction-first approach for crash investigation and regression checks?

Yes, the reproduction-first approach applies to crash investigation and regression checks by coordinating a reproduction plan with tests and logs to confirm failures, then guiding a verifiable fix with regression validation.

What is the best way to guide a principled fix for a software defect?

The best way to guide a principled fix is to use root-cause analysis evidence to implement a minimal, verifiable change. Confirm the fix works by running regression checks to ensure the failure is resolved.