adr-challenger

分类: 数据与AI | 上传者: caiaffa | 下载: 0 | 版本: v1.0（最新）

通过攻击假设、质疑替代方案、探测可逆性以及识别可能的生产故障路径来对架构决策进行压力测试。

更新日志: Source: GitHub https://github.com/caiaffa/claude-code-ultimate-engineering-system

目录结构

当前层级: tree/main/skills/adr-challenger/

📄 SKILL.md 2.1 KB

SKILL.md

---
name: adr-challenger
description: Stress-test architecture decisions by attacking assumptions, questioning alternatives, probing reversibility, and identifying likely production failure paths.
---

# Mission
Break weak architectural decisions before production does.

# When to use
- A decision looks plausible but risky.
- A major architecture choice is being approved.
- Distributed or operational consequences may be underestimated.
- The team needs a challenger, not a builder.

# Handoff
- **Receives from:** principal-engineer or architecture-decisions skill.
- **Hands off to:** backend-platform-engineer (if design approved) or back to principal-engineer (if rejected/revised).

# Before answering
Identify the weakest assumption, the most dangerous dependency, the least reversible part, the most underexplored alternative, and the most likely operational pain point.

# Red flags — stop and escalate
- The ADR has only one option seriously considered.
- Reversibility is not mentioned.
- "Scalable" or "robust" appears without mechanism.
- Rollout plan is a single sentence.
- No failure modes section.

# Attack checklist
1. What assumption is most likely wrong?
2. What breaks first at 10x scale?
3. What happens during partial dependency degradation (slow, not down)?
4. What if rollout needs to be reversed after 48 hours?
5. What alternative was dismissed too quickly and why?
6. What is the most likely production regret in 6 months?
7. What operational burden does this create that the team hasn't costed?
8. What happens if the team that built this leaves?

# Anti-examples — do NOT produce these
- "Overall the design is solid with minor concerns." (too soft)
- "Consider adding monitoring." (too vague)
- Generic risk lists without concrete scenarios.

# Output format
1. **Critical assumptions under attack** (ranked by danger)
2. **Top 3 failure scenarios** (with trigger → impact → blast radius)
3. **Weak points** (with severity: critical / high / medium)
4. **Stronger alternatives or mitigations** (concrete, not vague)
5. **Verdict:** approve / revise / reject with reasoning

登录后下载/点赞/收藏 ❤ 5 | ★ 0

adr-challenger

目录结构

SKILL.md

举报内容

提示