Verification before completion

Skill aaione/superpowers-zh/skills/verification-before-completion

当准备声称工作已完成、修复或通过时使用，在 commit 或创建 PR 之前 - 需要运行验证命令并在做出任何成功声明之前确认输出；永远是先证据后断言From its SKILL.md

Install

npx -y skills add aaione/superpowers-zh --skill verification-before-completion

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

8 stars8 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

4.1 KB, as published. Nobody here has run it

Verification Before Completion (完成前验证)

概述 (Overview)

未经验证就声称工作完成是不诚实，而不是效率。

核心原则: 证据先于声明，永远。

违反此规则的字面意思就是违反此规则的精神。

铁律 (The Iron Law)

没有新鲜的验证证据就没有完成声明 (NO COMPLETION CLAIMS WITHOUT FRESH VERIFICATION EVIDENCE)

如果你没有在此消息中运行验证命令，你不能声称它通过。

Gate Function (把关函数)

在声称任何状态或表达满意之前:

1. 识别: 什么命令证明此声明？
2. 运行: 执行完整命令 (fresh, complete)
3. 阅读: 完整输出，检查退出代码，计算失败
4. 验证: 输出是否确认声明？
   - 如果 NO: 带着证据陈述实际状态
   - 如果 YES: 带着证据陈述声明
5. 仅在此时: 做出声明

跳过任何步骤 = 撒谎，不是验证

常见失败 (Common Failures)

声明	需要	不足以
Tests pass	Test command output: 0 failures	之前的运行, "should pass"
Linter clean	Linter output: 0 errors	部分检查, 推断
Build succeeds	Build command: exit 0	Linter passing, logs look good
Bug fixed	Test original symptom: passes	代码已更改, 假定已修复
Regression test works	Red-green cycle verified	Test passes once
Agent completed	VCS diff shows changes	Agent 报告 "success"
Requirements met	Line-by-line checklist	Tests passing

危险信号 - 停止

使用 "should", "probably", "seems to"
在验证之前表达满意 ("Great!", "Perfect!", "Done!", etc.)
准备 commit/push/PR 而未经验证
信任 agent 成功报告
依赖部分验证
想着 "just this once"
累了想结束工作
任何暗示成功而没有运行验证的措辞

合理化预防 (Rationalization Prevention)

借口	现实
"现在应该能用了"	运行验证
"我很自信"	自信 ≠ 证据
"就这一次"	无例外
"Linter 通过了"	Linter ≠ 编译器
"Agent 说成功了"	独立验证
"我累了"	疲惫 ≠ 借口
"部分检查足够了"	部分证明不了什么
"措辞不同所以规则不适用"	精神高于字面

关键模式 (Key Patterns)

Tests:

✅ [Run test command] [See: 34/34 pass] "All tests pass"
❌ "Should pass now" / "Looks correct"

Regression tests (TDD Red-Green):

✅ Write → Run (pass) → Revert fix → Run (MUST FAIL) → Restore → Run (pass)
❌ "I've written a regression test" (without red-green verification)

Build:

✅ [Run build] [See: exit 0] "Build passes"
❌ "Linter passed" (linter doesn't check compilation)

Requirements:

✅ Re-read plan → Create checklist → Verify each → Report gaps or completion
❌ "Tests pass, phase complete"

Agent delegation:

✅ Agent reports success → Check VCS diff → Verify changes → Report actual state
❌ Trust agent report

为什么这很重要

来自 24 个失败记忆:

你的人类伙伴说 "I don't believe you" - 信任破裂
发布的未定义函数 - 会崩溃
发布的缺失需求 - 不完整的功能
浪费在虚假完成上的时间 → 重定向 → 返工
违反: "Honesty is a core value. If you lie, you'll be replaced."

何时应用 (When To Apply)

总是 (ALWAYS) 在之前:

成功/完成声明的任何变体
满意的任何表达
关于工作状态的任何积极陈述
Committing, PR creation, task completion
移动到下一个任务
委托给 agents

规则适用于:

确切的短语
释义和同义词
成功的暗示
任何暗示完成/正确性的沟通

总结 (The Bottom Line)

验证没有捷径。

运行命令。阅读输出。然后声称结果。

这是不可谈判的。