with-anti-rationalization

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

Anti-Rationalization Enforcement Skill

反合理化强制技能

Operator Context

操作场景

This skill operates as a composable modifier that wraps any task with explicit anti-rationalization enforcement. It implements the Gate Enforcement architectural pattern -- every phase transition requires evidence, every completion claim requires proof -- with Pressure Resistance embedded to prevent quality erosion under time or social pressure.

本技能作为可组合修饰器，为任意任务添加明确的反合理化强制机制。它采用Gate Enforcement（关卡强制）架构模式——每个阶段过渡都需要证据支持，每个完成声明都需要证明——同时嵌入Pressure Resistance（抗压能力），防止在时间或社交压力下出现质量下滑。

Hardcoded Behaviors (Always Apply)

硬编码行为（始终生效）

CLAUDE.md Compliance: Read and follow repository CLAUDE.md before starting any work
Full Pattern Loading: Load ALL anti-rationalization patterns, not just a relevant subset
Gate Enforcement: Every phase transition requires explicit evidence before proceeding
Pressure Resistance: Cannot skip verification even when user pressures for speed
Evidence-Based Completion: "Done" requires proof, not claims or confidence
Self-Check Protocol: Run the anti-rationalization self-check before marking any task complete

CLAUDE.md 合规性：开始任何工作前，阅读并遵循仓库中的CLAUDE.md
完整模式加载：加载所有反合理化模式，而非仅相关子集
关卡强制：每个阶段过渡都需要明确的证据才能继续
抗压能力：即使用户要求加快速度，也不能跳过验证步骤
基于证据的完成标准：“完成”需要证明，而非口头声明或主观信心
自我检查协议：在标记任务完成前，运行反合理化自我检查

Default Behaviors (ON unless disabled)

默认行为（默认开启，可关闭）

Verbose Check Output: Log each gate check and its pass/fail result visibly
Domain Pattern Loading: Load domain-specific anti-rationalization patterns (security, testing, review) when relevant to the task
Escalation on Ambiguity: Stop and ask user when requirements are unclear rather than guessing
Rationalization Logging: Document any detected rationalization attempts during execution
Completion Checklist: Run full verification checklist before declaring done

详细检查输出：清晰记录每个关卡检查及其通过/失败结果
领域模式加载：当任务涉及相关领域时，加载特定领域的反合理化模式（安全、测试、审查）
歧义时升级处理：当需求不明确时，停止操作并询问用户，而非自行猜测
合理化记录：记录执行过程中检测到的任何合理化尝试
完成检查清单：在宣布完成前运行完整的验证检查清单

Optional Behaviors (OFF unless enabled)

可选行为（默认关闭，可开启）

Audit Trail: Write detailed execution log of all gate checks to a file
Strict Mode: Treat warnings as failures at gate checks
Regression Verification: Run full test suite at every phase transition, not just at completion

审计追踪：将所有关卡检查的详细执行日志写入文件
严格模式：在关卡检查中将警告视为失败
回归验证：在每个阶段过渡时运行完整测试套件，而非仅在完成时运行

What This Skill CAN Do

本技能可实现的功能

Wrap any task with explicit anti-rationalization enforcement
Load all shared anti-rationalization patterns (core, security, testing, review)
Enforce phase gates with evidence requirements at every transition
Resist pressure to skip steps while remaining professional
Detect and surface rationalization attempts during execution
Require proof of completion rather than accepting claims

为任意任务添加明确的反合理化强制机制
加载所有共享反合理化模式（核心、安全、测试、审查）
在每个阶段过渡时强制执行带证据要求的关卡检查
在保持专业性的同时，拒绝跳过步骤的要求
检测并呈现执行过程中的合理化尝试
要求提供完成证明，而非接受口头声明

What This Skill CANNOT Do

本技能无法实现的功能

Replace domain-specific skills (debugging, refactoring, testing have their own methodologies)
Make tasks faster -- this adds overhead deliberately for safety
Override explicit user decisions about scope or technical approach
Guarantee zero defects -- rigor reduces risk, it does not eliminate it
Operate without an underlying task -- this is a modifier, not a standalone workflow

替代特定领域技能（调试、重构、测试有各自的方法论）
加快任务速度——为了安全性，本技能会刻意增加开销
覆盖用户关于范围或技术方案的明确决策
保证零缺陷——严谨性降低风险，但无法消除风险
在无基础任务的情况下运行——本技能是修饰器，而非独立工作流

Instructions

操作说明

Phase 1: LOAD PATTERNS

阶段1：加载模式

Goal: Load all anti-rationalization patterns relevant to the task before starting work.

Step 1: Identify task domain

Classify the task to determine which domain-specific patterns apply:

Domain	Pattern to Load
Any task	`anti-rationalization-core.md`
Code review	`anti-rationalization-review.md`
Testing	`anti-rationalization-testing.md`
Security	`anti-rationalization-security.md`
Multi-phase work	`gate-enforcement.md`
User pressure detected	`pressure-resistance.md`
Pre-completion	`verification-checklist.md`

Step 2: Load and acknowledge patterns

Read the identified shared-pattern files. Internalize the rationalization tables and enforcement rules. State which patterns were loaded and why -- this creates accountability.

Gate: All relevant patterns loaded and acknowledged. Proceed only when gate passes.

目标：在开始工作前，加载与任务相关的所有反合理化模式。

步骤1：识别任务领域

对任务进行分类，确定适用的特定领域模式：

任务领域	需加载的模式
任意任务	`anti-rationalization-core.md`
代码审查	`anti-rationalization-review.md`
测试	`anti-rationalization-testing.md`
安全	`anti-rationalization-security.md`
多阶段工作	`gate-enforcement.md`
检测到用户压力	`pressure-resistance.md`
完成前	`verification-checklist.md`

步骤2：加载并确认模式

读取已识别的共享模式文件，内化合理化表格和强制规则。说明加载了哪些模式及原因——这会建立问责制。

关卡：所有相关模式已加载并确认。仅当关卡通过后才能继续。

Phase 2: EXECUTE WITH ENFORCEMENT

阶段2：强制执行任务

Goal: Run the underlying task with anti-rationalization checks at every transition.

Step 1: Delegate to appropriate methodology

This skill wraps other skills. If the task involves debugging, follow systematic-debugging methodology. If refactoring, follow systematic-refactoring. The anti-rationalization layer adds checks on top.

Step 2: At each phase transition, run gate check

For each transition, verify: (1) all exit criteria met, (2) evidence documented not just claimed, (3) anti-rationalization table reviewed, (4) no rationalization detected. Then run a rationalization scan -- am I assuming without verifying? Skipping because it "looks right"? Rushing from perceived pressure? If any answer is YES: STOP and address the rationalization before proceeding.

Step 3: Handle pressure resistance

If the user requests skipping a step:

Acknowledge the request
Explain why the step matters (one sentence)
Proceed with the step
If user insists on a non-security matter, note the risk and comply

Gate: Task phases executed with all gate checks passing. Proceed only when gate passes.

目标：在每个阶段过渡时进行反合理化检查，运行基础任务。

步骤1：委托给合适的方法论

本技能用于包装其他技能。如果任务涉及调试，请遵循系统调试方法论；如果是重构，请遵循系统重构方法论。反合理化层会在其基础上添加检查。

步骤2：在每个阶段过渡时运行关卡检查

对于每个过渡阶段，验证：(1) 所有退出标准已满足，(2) 证据已记录而非仅声明，(3) 已审查反合理化表格，(4) 未检测到合理化行为。然后运行合理化扫描——我是验证了还是只是假设？是否因为“看起来正确”而跳过步骤？是否因感知到压力而匆忙行事？如果任何答案为是：停止操作并解决该合理化问题后再继续。

步骤3：处理抗压情况

如果用户要求跳过步骤：

确认收到请求
用一句话解释该步骤的重要性
继续执行该步骤
如果用户在非安全问题上坚持，记录风险并遵守要求

关卡：任务各阶段已执行且所有关卡检查通过。仅当关卡通过后才能继续。

Phase 3: VERIFY WITH FULL CHECKLIST

阶段3：使用完整检查清单验证

Goal: Verify completion with the full verification checklist and anti-rationalization self-check.

Step 1: Run verification checklist

Check	Verified?	Evidence
All stated requirements addressed	[ ]	[specific evidence]
Tests pass (if applicable)	[ ]	[test output]
No regressions introduced	[ ]	[existing test output]
Error handling in place	[ ]	[error paths tested]
Code compiles/lints	[ ]	[build output]
Anti-rationalization table reviewed	[ ]	[self-check completed]

Step 2: Run completion self-check

markdown

undefined

目标：使用完整的验证检查清单和反合理化自我检查来验证完成情况。

步骤1：运行验证检查清单

检查项	是否已验证？	证据
所有指定需求已处理	[ ]	[具体证据]
测试通过（如适用）	[ ]	[测试输出]
未引入回归问题	[ ]	[现有测试输出]
已设置错误处理	[ ]	[已测试错误路径]
代码可编译/通过代码检查	[ ]	[构建输出]
已审查反合理化表格	[ ]	[已完成自我检查]

步骤2：运行完成自我检查

markdown

undefined

Completion Self-Check

Did I verify or just assume?
Did I run tests or just check code visually?
Did I complete everything or just the "important" parts?
Would I bet $100 this works correctly?
Can I show evidence (output, test results)?


If ANY answer is uncertain, return to Phase 2 and address the gap.

**Step 3: Document completion evidence**

Summarize: task description, patterns loaded, gate checks passed, rationalizations detected and addressed, and final evidence proving the task is complete.

**Gate**: All verification steps pass. Self-check is clean. Evidence documented. Task is complete.

---

Did I verify or just assume?
Did I run tests or just check code visually?
Did I complete everything or just the "important" parts?
Would I bet $100 this works correctly?
Can I show evidence (output, test results)?


如果任何答案不确定，返回阶段2并解决差距。

**步骤3：记录完成证据**

总结：任务描述、加载的模式、通过的关卡检查、检测到并解决的合理化行为，以及证明任务已完成的最终证据。

**关卡**：所有验证步骤通过。自我检查无问题。证据已记录。任务完成。

---

Examples

示例

Example 1: Critical Production Change

示例1：关键生产变更

User says: "/with-anti-rationalization deploy the payment processor update" Actions:

Load core + security anti-rationalization patterns (LOAD)
Execute deployment with gate checks at each phase (EXECUTE)
Resist any "just ship it" pressure with evidence requirements
Full verification checklist before declaring done (VERIFY) Result: Deployment verified with evidence at every step

用户说："/with-anti-rationalization 部署支付处理器更新" 操作：

加载核心+安全反合理化模式（加载阶段）
在每个阶段执行部署并进行关卡检查（执行阶段）
以证据要求拒绝任何“直接发布”的压力
在宣布完成前运行完整验证检查清单（验证阶段）结果：每个步骤都有证据验证的部署

Example 2: Security-Sensitive Code Review

示例2：安全敏感型代码审查

User says: "/with-anti-rationalization review the authentication module" Actions:

Load core + security + review anti-rationalization patterns (LOAD)
Review with explicit checks against "internal only" and "low risk" rationalizations (EXECUTE)
Every finding documented with evidence, not dismissed
Completion self-check confirms no findings were skipped (VERIFY) Result: Thorough review with no rationalized dismissals

用户说："/with-anti-rationalization 审查认证模块" 操作：

加载核心+安全+审查反合理化模式（加载阶段）
针对“仅限内部”和“低风险”的合理化行为进行明确检查（执行阶段）
所有发现都有证据记录，未被忽略
完成自我检查确认未跳过任何发现（验证阶段）结果：彻底的审查，无合理化忽略的情况

Error Handling

错误处理

Error: "Pattern File Not Found"

错误："Pattern File Not Found"

Cause: Shared pattern file missing or path changed Solution:

Check
```
skills/shared-patterns/
```
for available files
If file was renamed, use the new name
If file was deleted, apply the core patterns from CLAUDE.md as fallback
Document which pattern could not be loaded

原因：共享模式文件缺失或路径更改解决方案：

检查
```
skills/shared-patterns/
```
目录下的可用文件
如果文件已重命名，使用新名称
如果文件已删除，使用CLAUDE.md中的核心模式作为备选
记录无法加载的模式

Error: "Gate Check Fails Repeatedly"

错误："Gate Check Fails Repeatedly"

Cause: Task requirements unclear, or task is fundamentally blocked Solution:

Re-read the gate criteria -- are they appropriate for this task?
If requirements are unclear, escalate to user for clarification
If technically blocked, document the blocker and ask user how to proceed
Do NOT weaken the gate criteria to force a pass

原因：任务需求不明确，或任务根本受阻解决方案：

重新阅读关卡标准——是否适用于此任务？
如果需求不明确，升级到用户处寻求澄清
如果技术上受阻，记录障碍并询问用户如何继续
不要为了强行通过而削弱关卡标准

Error: "User Insists on Skipping Verification"

错误："User Insists on Skipping Verification"

Cause: Time pressure, frustration, or genuine scope reduction Solution:

Distinguish quality skip (resist) from scope preference (respect)
If quality: explain risk once, note risk in output, comply if user insists again
If security: refuse and explain -- security shortcuts are non-negotiable
Document that verification was skipped at user request

原因：时间压力、挫折感，或真正的范围缩减解决方案：

区分质量跳过（拒绝）和范围偏好（尊重）
如果是质量问题：解释一次风险，在输出中记录风险，如果用户再次坚持则遵守
如果是安全问题：拒绝并解释——安全捷径是不可协商的
记录验证是在用户要求下跳过的

Anti-Patterns

反模式

Anti-Pattern 1: Performative Checking

反模式1：形式化检查

What it looks like: Running gate checks but rubber-stamping them all as PASS without reading evidence Why wrong: Gate checks that always pass provide zero value. The check is the evidence review, not the checkbox. Do instead: Read the evidence for each criterion. If you cannot articulate WHY it passes, it does not pass.

表现：运行关卡检查但未经阅读证据就全部标记为通过 错误原因：总是通过的关卡检查毫无价值。检查的核心是证据审查，而非勾选框。 正确做法：阅读每个标准的证据。如果无法说明通过的原因，则视为未通过。

Anti-Pattern 2: Rationalization Laundering

反模式2：合理化洗白

What it looks like: Reframing a skipped step as "not applicable" rather than "skipped" Why wrong: "Not applicable" is sometimes legitimate, but it is also the most common way to rationalize skipping steps. Do instead: For every "N/A" judgment, state WHY it does not apply. If the reason is weak, do the step.

表现：将跳过的步骤重新定义为“不适用”而非“已跳过” 错误原因：“不适用”有时是合理的，但也是跳过步骤最常见的合理化方式。 正确做法：对于每个“不适用”的判断，说明为什么不适用。如果理由不充分，则执行该步骤。

Anti-Pattern 3: Selective Pattern Loading

反模式3：选择性加载模式

What it looks like: Loading only anti-rationalization-core and skipping domain-specific patterns Why wrong: Domain-specific patterns catch domain-specific rationalizations that the core misses. Do instead: Classify the task domain in Phase 1 and load ALL matching patterns.

表现：仅加载核心反合理化模式，跳过特定领域模式 错误原因：特定领域模式能捕捉核心模式遗漏的特定领域合理化行为。 正确做法：在阶段1中对任务领域进行分类，并加载所有匹配的模式。

Anti-Pattern 4: Pressure Capitulation

反模式4：屈服于压力

What it looks like: Immediately dropping verification when user says "just do it" Why wrong: The entire purpose of this skill is to resist shortcuts. Immediate capitulation defeats the purpose. Do instead: Follow the pressure resistance framework: acknowledge, explain, proceed. Comply only after explaining risk.

表现：当用户说“直接做”时立即放弃验证 错误原因：本技能的全部目的是拒绝走捷径。立即屈服会失去其意义。 正确做法：遵循抗压框架：确认、解释、执行。仅在解释风险后再遵守。

Anti-Pattern 5: Anti-Rationalization Theater

反模式5：反合理化形式主义

What it looks like: Spending more time on the checking framework than on the actual task Why wrong: The goal is correct output, not elaborate process documentation. Checks should be proportionate. Do instead: Scale check depth to task risk. Critical production changes get full ceremony. A three-file refactor gets lighter gates.

表现：在检查框架上花费的时间比实际任务更多 错误原因：目标是正确的输出，而非复杂的流程文档。检查应与任务风险相匹配。 正确做法：根据任务风险调整检查深度。关键生产变更需要完整的流程，三文件重构则需要更简化的关卡。

References

参考资料

This skill composes these shared patterns:

Anti-Rationalization Core - Universal rationalization detection and prevention
Anti-Rationalization Review - Review-specific patterns
Anti-Rationalization Testing - Testing-specific patterns
Anti-Rationalization Security - Security-specific patterns
Gate Enforcement - Phase transition enforcement
Pressure Resistance - Handling pushback professionally
Verification Checklist - Pre-completion verification

本技能组合了以下共享模式：

Anti-Rationalization Core - 通用合理化检测与预防
Anti-Rationalization Review - 审查特定模式
Anti-Rationalization Testing - 测试特定模式
Anti-Rationalization Security - 安全特定模式
Gate Enforcement - 阶段过渡强制
Pressure Resistance - 专业处理反对意见
Verification Checklist - 完成前验证

Domain-Specific Anti-Rationalization

特定领域反合理化

Rationalization	Why It's Wrong	Required Action
"I loaded the patterns, that's enough"	Loading is not applying	Actively check against patterns at each gate
"This task is simple, full rigor is overkill"	Simplicity assessment is itself a rationalization risk	Apply proportionate rigor, but never zero
"User seems frustrated, I'll ease up"	Frustration does not change correctness requirements	Acknowledge frustration, maintain standards
"The gate basically passes"	Basically is not actually	Either it passes with evidence or it does not

合理化行为	错误原因	要求的操作
"我已经加载了模式，这就够了"	加载不等于应用	在每个关卡主动对照模式进行检查
"这个任务很简单，完全严谨是多余的"	对简单性的评估本身就存在合理化风险	应用与风险匹配的严谨性，但绝不完全不应用
"用户看起来很沮丧，我会放宽标准"	沮丧不会改变正确性要求	承认沮丧情绪，维持标准
"关卡基本通过"	基本通过不等于实际通过	要么有证据证明通过，要么未通过