AgentSkillsCN

ci-debugging

通过分析日志、配置文件以及环境差异，系统性地排查 CI/CD 管道故障。

SKILL.md
--- frontmatter
name: ci-debugging
description: Systematically debug CI/CD pipeline failures by analyzing logs, configuration, and environment differences.
version: 1.0.0
tier: 0
license: MIT

CI Debugging

Use this skill when a CI/CD pipeline fails and the cause is not immediately obvious.

When to Trigger

  • •A CI job fails after a code change.
  • •A previously passing pipeline starts failing.
  • •CI behavior differs from local (inner loop) results.

Workflow

  1. •Read the failure log. Start from the bottom — the root cause is usually near the end.
  2. •Classify the failure:
    • •Build failure — compilation, missing dependencies, version mismatch
    • •Test failure — failing assertions, timeout, flaky test
    • •Lint/format failure — style violations, type errors
    • •Infrastructure failure — network, permissions, resource limits, runner issues
    • •Configuration failure — env vars, secrets, pipeline config syntax
  3. •Check for environment differences. Compare CI environment against local:
    • •Language/runtime versions
    • •Environment variables (especially secrets availability)
    • •OS and architecture
    • •Installed tooling versions
  4. •Reproduce locally if possible. Use the same commands CI runs.
  5. •Fix the root cause. Do not mask failures with retries or skips unless the failure is genuinely flaky and documented.
  6. •Verify the fix passes in CI before considering the task complete.

Anti-Patterns

  • •Do not add continue-on-error or || true to silence failures.
  • •Do not skip tests that fail in CI but pass locally — find the environment difference.
  • •Do not retry without understanding why it failed the first time.
  • •Do not modify CI config without understanding the pipeline structure.

Quality Checks

  • • Root cause identified (not just symptom)
  • • Fix verified in CI (not just locally)
  • • No tests skipped or silenced
  • • CI config changes explained in commit message