zhenxuanshi-ship-it

business-doc-integrity

0
0
# Install this skill:
npx skills add zhenxuanshi-ship-it/business-doc-skill-suite --skill "business-doc-integrity"

Install specific skill from multi-skill repository

# Description

Check business and product documents for consistency, internal logic, terminology stability, unsupported claims, cross-section contradictions, and execution credibility. Use when the user wants to verify a 项目可行性研究报告, 项目建议书, PRD, 产品需求说明书, 架构设计, 技术方案, 实施方案, or final draft before review, approval, or delivery. Triggers on requests like: 检查一致性, 做文档校验, 看看有没有前后冲突, 查查这份方案有没有硬伤, 核查这份 PRD, 校验架构设计, integrity check this document, consistency check this proposal, validate this draft.

# SKILL.md


name: business-doc-integrity
description: Check business and product documents for consistency, internal logic, terminology stability, unsupported claims, cross-section contradictions, and execution credibility. Use when the user wants to verify a 项目可行性研究报告, 项目建议书, PRD, 产品需求说明书, 架构设计, 技术方案, 实施方案, or final draft before review, approval, or delivery. Triggers on requests like: 检查一致性, 做文档校验, 看看有没有前后冲突, 查查这份方案有没有硬伤, 核查这份 PRD, 校验架构设计, integrity check this document, consistency check this proposal, validate this draft.


Business Doc Integrity

Check whether a business document is internally trustworthy and decision-ready.

This is not academic citation verification.
This skill focuses on document integrity in enterprise writing:
- consistency
- logic
- terminology
- scope alignment
- numeric stability
- execution realism
- structure-to-content balance

Core rule

Do not only point out that something feels vague.

Whenever possible, identify:
1. what is inconsistent,
2. where it appears,
3. why it is risky,
4. what should be reconciled.

Main check modes

1. full

Use for a complete integrity pass.

Check:
- terminology consistency
- number consistency
- timeline consistency
- scope consistency
- section-to-section logic
- recommendation support
- risk closure

2. consistency-only

Use when the user mainly wants to find contradictions.

3. logic-only

Use when the user mainly wants to test whether conclusions and recommendations are supported.

4. final-gate

Use before approval, submission, or final delivery.

This mode is stricter and should answer:
- is this ready to circulate?
- what still blocks decision confidence?

What to check

A. Terminology consistency

Look for:
- one thing called by multiple names
- module names changing across sections
- role names shifting
- business terms and technical terms being mixed carelessly

Examples:
- “统一认证中心” later becomes “SSO平台” and later “认证网关” without clarification
- “试点阶段” and “一期建设” used as if they mean the same thing

B. Number consistency

Look for:
- budget numbers changing between sections
- timeline counts not matching milestones
- user volume / traffic / resource assumptions changing without explanation
- counts in summary not matching detail tables

C. Timeline consistency

Look for:
- milestone order conflicts
- implementation tasks happening before prerequisites
- rollout timing inconsistent with dependency timing

D. Scope consistency

Look for:
- scope and non-scope contradicting each other
- summary promising more than detailed sections support
- PRD requirements exceeding the declared objective
- architecture covering less than the claimed business scope

E. Logic and recommendation support

Look for:
- recommendation appears before real comparison
- conclusion is stronger than evidence
- “necessary” is asserted but not shown
- “feasible” is declared without addressing key constraints
- budget strength exceeds budget basis
- implementation commitment exceeds implementation preparation

F. Risk closure

Look for:
- major risks named but not mitigated
- mitigation measures too generic to be actionable
- no owner or trigger condition for risk handling
- dependencies mentioned without contingency thinking

G. Cross-document alignment

When the user provides multiple related materials, check alignment between them:
- brief vs proposal
- PRD vs architecture design
- proposal vs implementation plan
- feasibility report vs executive summary

H. Skeletonization / hollow-outline risk

Look for:
- too many headings with too little development under each
- sections that only restate the heading in sentence form
- core claims appearing in summary/conclusion without being developed earlier
- multiple sections with very low information gain
- documents that look complete structurally but remain thin operationally

I. Insufficient-support risk

Look for:
- recommendation is stronger than the body of analysis
- budget numbers are more specific than their basis
- scope is more ambitious than the implementation path supports
- conclusions rely on sections that were only introduced, not developed
- key claims are present, but the document does not spend enough text actually proving them

Workflow

Step 1: Identify document type

Use references/integrity-matrix.md to adjust emphasis.

Step 2: Scan for structural weak points

Before line-by-line issues, detect the big failure modes:
- recommendation unsupported
- scope drifting
- architecture-document mismatch
- optimistic timeline fantasy
- decorative risk section
- hollow-outline structure with insufficient expansion
- long-form document that still reads like a compressed outline
- budget / recommendation / implementation sections that are thinner than their importance requires

Step 3: Produce issue list with evidence

For each issue, include:
- issue title
- severity
- location
- conflicting statements or sections
- why it matters
- suggested reconciliation

Step 4: Give gate decision

Use one of these:
- Pass
- Pass with fixes
- Needs revision before circulation
- Not credible enough yet

Severity scale

  • Critical: breaks trust or blocks approval/implementation
  • High: likely to cause major confusion, rework, or bad decisions
  • Medium: important inconsistency or weakness, but repairable
  • Low: polish-level issue

Output format

1. Integrity summary

  • what is stable
  • what is unstable
  • whether the document is safe to circulate

2. Key findings

List the most important contradictions or unsupported jumps first.
Also call out if the document is structurally complete but informationally thin.

3. Detailed issue table

Include:
- issue
- severity
- location
- conflicting or weak statements
- suggested fix

4. Reconciliation checklist

Group into:
- Must reconcile now
- Should tighten next
- Optional cleanup

5. Gate decision

State the final gate result clearly.

Good integrity feedback

Good:
- “Section 2 says Phase 1 only covers internal users, but Section 4.3 includes partner-side portal access. Scope needs reconciliation.”
- “The proposal recommends immediate rollout, but the implementation section still depends on data governance rules that are marked as pending. Recommendation is ahead of prerequisites.”

Bad:
- “This feels a little inconsistent in places.”

References

Read when relevant:
- references/integrity-matrix.md — integrity focus by document type
- references/failure-patterns.md — common business document failure modes
- references/output-format.md — recommended report structure

# Supported AI Coding Agents

This skill is compatible with the SKILL.md standard and works with all major AI coding agents:

Learn more about the SKILL.md standard and how to use these skills with your preferred AI coding agent.