AgentSkillsCN

debug-triage

采用“快速重现→隔离→最小化失败测试→修复→验证”的流程，高效调试并排查测试失败、回归问题、不稳定行为，或生产环境中的各类故障。当您需要调试代码、调查 CI 失败，或在任意仓库中最大限度降低回归风险时，本指南将助您事半功倍。

SKILL.md
--- frontmatter
name: debug-triage
description: "Debug and triage failing tests, regressions, flaky behavior, or production issues with a fast repro -> isolate -> minimal failing test -> fix -> verify workflow. Use when debugging, investigating CI failures, or minimizing regression risk in any repo."

Debug Triage

Overview

Apply a consistent, minimal-change debugging workflow that yields a reproducible failure, a focused fix, and verification tests.

Workflow

1) Capture the symptom

  • •Record the error, stack trace, and environment details.
  • •Identify the first failing test or observable failure.

2) Reproduce locally

  • •Re-run the failing test or command.
  • •If flaky, run multiple times and note variability.
  • •If no repro, ask for logs, seeds, config, or CI artifacts.

3) Isolate the cause

  • •Reduce input and scope to the smallest failing case.
  • •Narrow the failure by bisecting or commenting out changes.
  • •Add temporary logs or assertions to confirm hypotheses.

4) Lock the bug with a test

  • •Write the smallest failing test that reproduces the bug.
  • •Ensure it fails before fixing (Red).

5) Fix minimally

  • •Make the smallest change to pass the test (Green).
  • •Refactor only after green tests.

6) Verify

  • •Run the minimal test set and any higher-level tests if risk is high.
  • •Remove temporary logs and confirm cleanup.

Output template (adapt as needed)

  • •Root cause: <short explanation>
  • •Fix: <what changed>
  • •Tests run: <commands and results>
  • •Risks/notes: <follow-ups or gaps>