使用方法
- 复制安装命令,在自己的智能体环境安装前先检查源码包。
- 从原始文档核对所需工具、凭据和支持的宿主。
- 说明输入材料与期望交付物,并对照来源检查输出。
npx skills add https://github.com/alexbruno/dev-skills --skill systematic-debugging开始前需了解
- 安装不包含模型订阅、API 额度或访问文件的权限。
- 仓库星数和安装量反映活跃情况,不等于独立质量或安全测试。
按源仓库内容呈现,保留标题、案例、代码、表格、链接以及原文引用的演示图片。
Systematic Debugging
Find and prove the root cause before changing any production code. Never guess, apply shotgun patches, or silence errors to make symptoms disappear.
Core Rule
No fix without a confirmed root cause. Every bug fix must follow: Reproduce -> Isolate -> Prove Root Cause -> Regression Test -> Fix -> Verify.
Debugging Process
1. Reproduce Reliably
- Capture exact reproduction steps, inputs, environment conditions, and expected vs. actual behavior.
- If the failure cannot be reproduced yet, gather logs, traces, or state snapshots first—do not speculate.
2. Isolate & Form Hypotheses
- Narrow the failure boundary (recent changes via
git diff/git log, failing inputs, or call paths viacodebase-indexingif active). - Form one concrete, testable hypothesis at a time about why the failure occurs.
- Test the hypothesis with targeted observation (logs, minimal reproduction, or debugger/test run).
3. Confirm & Document Root Cause
- Confirm the exact mechanism causing the bug with direct evidence.
- If
workflow/SPEC.mdis active for this bug fix, record the reproduction steps and confirmed root cause there before planning the fix.
4. Write a Failing Regression Test First
- Create an automated test (or reproducible check) that fails for the exact reason of the bug.
- Verify it fails before applying the fix so you know the test actually catches the regression.
5. Fix the Root Cause & Verify
- Apply the minimal, clean fix that addresses the root cause—not a workaround around the symptom.
- Run the regression test and the full relevant test suite to confirm no regressions were introduced.
Anti-Patterns to Avoid
- Shotgun debugging: Changing multiple unrelated things at once hoping one works.
- Symptom masking: Catching and ignoring exceptions, adding arbitrary retries/sleeps, or loosening assertions to pass tests.
- Speculative fixes: Editing code before you can explain why it failed.