name: req-review-risk
description: >
Requirements review risk screening. Analyzes requirement documents for ambiguity,
logic conflicts, missing constraints, unclear boundaries, missing compatibility
requirements and other risks. Outputs a structured Markdown review checklist with
severity labels (P0/P1/P2) and category tags, AND auto-generates a standalone HTML
report. Use when user pastes requirement text for review, asks for 需求评审,
requirement quality check, risk screening, or mentions catching requirement issues
before development.
需求评审风险筛查
Analyze requirement text, output a labeled Markdown checklist AND an HTML report.
Risk Categories
Label every finding with exactly one category tag:
| Tag |
Category |
[模糊] |
Ambiguity - vague wording, multiple interpretations |
[冲突] |
Logic Conflict - contradictory requirements |
[缺失] |
Missing Constraint - no perf target, no data range, no error handling |
[边界] |
Unclear Boundary - scope unclear, missing interface definition |
[兼容] |
Missing Compatibility - browser/OS/version not specified |
[安全] |
Security Gap - auth, encryption, input validation missing |
[可测] |
Testability - not verifiable, no acceptance criteria |
Severity
- P0 阻断: Must resolve before dev. Causes rework or system failure.
- P1 重要: Should resolve. Significant issues or delays likely.
- P2 建议: Nice to resolve. Minor clarity improvement.
Process
- Read entire requirement text.
- Scan each statement against all 7 risk categories.
- For each finding, extract exact original text. One finding per row.
- Assign category tag, severity, issue description, and specific suggestion.
- Output a Markdown checklist in the chat.
- MUST also generate a standalone HTML report file (see below).
HTML Report Generation
Always generate an HTML file alongside the Markdown output.
Filename Convention
Name the file using the following priority order:
需求评审-{产品名}-V{版本号}-{YYYYMMDD}.html
Extraction rules:
- Scan the requirement text for explicit patterns:
XX系统 V2.0, XX模块, XX平台 V1, XX项目 3.0
- If found, use:
需求评审-{产品名}-V{版本号}-{YYYYMMDD}.html (e.g. 需求评审-订单系统-V2.1-20260609.html)
- If only product name found, omit version:
需求评审-{产品名}-{YYYYMMDD}.html
- If neither found, use the first meaningful noun from the text as product name
- Fallback:
需求评审-{YYYYMMDD}-{HHmm}.html
- Always save to current working directory.
HTML Template Rules
The HTML must be self-contained (inline CSS, no external deps) and include:
- Header: title, review date, document/product name
- 需求原文: original text in a styled blockquote
- 概览: summary grid with P0/P1/P2/total counts
- 风险清单: tables grouped by severity, columns: # | 类别 (tag) | 原文引用 | 问题描述 | 修改建议
- 评审结论: one-sentence assessment + prioritized action items
- Action bar: print/save PDF button, select-all copy button
- Print-friendly CSS (@media print)
Style guidelines: clean modern UI, blue-purple gradient header, color-coded tags per category, severity badges (P0=red, P1=orange, P2=blue).
Output Format (Markdown)
# 需求评审风险筛查报告
## 概览
| 指标 | 数量 |
|------|------|
| P0 阻断 | X |
| P1 重要 | X |
| P2 建议 | X |
| 合计 | X |
## 风险清单
### [P0] 阻断风险
| # | 类别 | 原文引用 | 问题描述 | 修改建议 |
|---|------|----------|----------|----------|
### [P1] 重要风险
...
### [P2] 建议优化
...
Rules
If the original requirement text exceeds 15 lines, wrap the blockquote in details/summary with line count (e.g. '点击展开需求原文(共28行)'), default collapsed. In print media, auto-expand.
Cite original text verbatim in 原文引用, do not paraphrase.
One finding per row. Multiple issues in one sentence = multiple rows.
Be specific in suggestions: state exactly what to add, not just "be clearer".
Skip empty severity sections.
Output Markdown first, then state the HTML file path.
Do not add closing remarks beyond the file path.
1---2name: req-review-risk3description: ---4---5---6name: req-review-risk7description: >8 Requirements review risk screening. Analyzes requirement documents for ambiguity,9 logic conflicts, missing constraints, unclear boundaries, missing compatibility10 requirements and other risks. Outputs a structured Markdown review checklist with11 severity labels (P0/P1/P2) and category tags, AND auto-generates a standalone HTML12 report. Use when user pastes requirement text for review, asks for 需求评审,13 requirement quality check, risk screening, or mentions catching requirement issues14 before development.15---1617# 需求评审风险筛查1819Analyze requirement text, output a labeled Markdown checklist AND an HTML report.2021## Risk Categories2223Label every finding with exactly one category tag:2425| Tag | Category |26|-----|----------|27| `[模糊]` | Ambiguity - vague wording, multiple interpretations |28| `[冲突]` | Logic Conflict - contradictory requirements |29| `[缺失]` | Missing Constraint - no perf target, no data range, no error handling |30| `[边界]` | Unclear Boundary - scope unclear, missing interface definition |31| `[兼容]` | Missing Compatibility - browser/OS/version not specified |32| `[安全]` | Security Gap - auth, encryption, input validation missing |33| `[可测]` | Testability - not verifiable, no acceptance criteria |3435## Severity3637- **P0 阻断**: Must resolve before dev. Causes rework or system failure.38- **P1 重要**: Should resolve. Significant issues or delays likely.39- **P2 建议**: Nice to resolve. Minor clarity improvement.4041## Process42431. Read entire requirement text.442. Scan each statement against all 7 risk categories.453. For each finding, extract **exact original text**. One finding per row.464. Assign category tag, severity, issue description, and specific suggestion.475. Output a Markdown checklist in the chat.486. **MUST** also generate a standalone HTML report file (see below).4950## HTML Report Generation5152Always generate an HTML file alongside the Markdown output.5354### Filename Convention5556Name the file using the following priority order:5758```59需求评审-{产品名}-V{版本号}-{YYYYMMDD}.html60```6162**Extraction rules:**63- Scan the requirement text for explicit patterns: `XX系统 V2.0`, `XX模块`, `XX平台 V1`, `XX项目 3.0`64- If found, use: `需求评审-{产品名}-V{版本号}-{YYYYMMDD}.html` (e.g. `需求评审-订单系统-V2.1-20260609.html`)65- If only product name found, omit version: `需求评审-{产品名}-{YYYYMMDD}.html`66- If neither found, use the first meaningful noun from the text as product name67- Fallback: `需求评审-{YYYYMMDD}-{HHmm}.html`68- Always save to current working directory.6970### HTML Template Rules7172The HTML must be self-contained (inline CSS, no external deps) and include:731. **Header**: title, review date, document/product name742. **需求原文**: original text in a styled blockquote753. **概览**: summary grid with P0/P1/P2/total counts764. **风险清单**: tables grouped by severity, columns: # | 类别 (tag) | 原文引用 | 问题描述 | 修改建议775. **评审结论**: one-sentence assessment + prioritized action items786. **Action bar**: print/save PDF button, select-all copy button797. Print-friendly CSS (@media print)8081Style guidelines: clean modern UI, blue-purple gradient header, color-coded tags per category, severity badges (P0=red, P1=orange, P2=blue).8283## Output Format (Markdown)8485```markdown86# 需求评审风险筛查报告8788## 概览89| 指标 | 数量 |90|------|------|91| P0 阻断 | X |92| P1 重要 | X |93| P2 建议 | X |94| 合计 | X |9596## 风险清单9798### [P0] 阻断风险99| # | 类别 | 原文引用 | 问题描述 | 修改建议 |100|---|------|----------|----------|----------|101102### [P1] 重要风险103...104105### [P2] 建议优化106...107```108109## Rules110111- If the original requirement text exceeds 15 lines, wrap the blockquote in details/summary with line count (e.g. '点击展开需求原文(共28行)'), default collapsed. In print media, auto-expand.112113114- Cite original text verbatim in `原文引用`, do not paraphrase.115- One finding per row. Multiple issues in one sentence = multiple rows.116- Be specific in suggestions: state exactly what to add, not just "be clearer".117- Skip empty severity sections.118- Output Markdown first, then state the HTML file path.119- Do not add closing remarks beyond the file path.