doc-prd-fixer（文档 PRD 修复程序）

doc-prd-fixer

Purpose

Automated fix skill that reads the latest audit/review report and applies fixes to PRD documents. This skill bridges the gap between doc-prd-reviewer/doc-prd-audit (which identify issues) and the corrected PRD, enabling iterative improvement cycles.

Layer: 2 (PRD Quality Improvement)

Upstream: PRD document, Audit/Review Report (PRD-NN.A_audit_report_vNNN.md preferred, PRD-NN.R_review_report_vNNN.md legacy), BRD (source requirements)

Downstream: Fixed PRD, Fix Report (PRD-NN.F_fix_report_vNNN.md)

When to Use This Skill

Use doc-prd-fixer when:

• After Review: Run after doc-prd-reviewer identifies issues
• Iterative Improvement: Part of Review -> Fix -> Review cycle
• Automated Pipeline: CI/CD integration for quality gates
• Batch Fixes: Apply fixes to multiple PRDs based on review reports

Do NOT use when:

• No audit/review report exists (run doc-prd-audit or doc-prd-reviewer first)
• Creating new PRD (use doc-prd or doc-prd-autopilot)
• Only need validation (use doc-prd-validator)

Skill Dependencies

Workflow Overview

flowchart TD
    A[Input: PRD Path] --> B[Find Latest Audit/Review Report]
    B --> C{Report Found?}
    C -->|No| D[Run doc-prd-audit or doc-prd-reviewer First]
    C -->|Yes| E[Parse Report]

    E --> F[Categorize Issues]

    subgraph FixPhases["Fix Phases"]
        F --> F0[Phase 0: Fix Structure Violations]
        F0 --> G[Phase 1: Create Missing Files]
        G --> H[Phase 2: Fix Broken Links]
        H --> I[Phase 3: Fix Element IDs]
        I --> J[Phase 4: Fix Content Issues]
        J --> K[Phase 5: Update References]
        K --> K2[Phase 6: Handle Upstream Drift]
    end

    K2 --> L[Write Fixed PRD]
    L --> M[Generate Fix Report]
    M --> N{Re-run Review?}
    N -->|Yes| O[Invoke doc-prd-reviewer]
    O --> P{Score >= Threshold?}
    P -->|No, iterations < max| F
    P -->|Yes| Q[COMPLETE]
    N -->|No| Q

Fix Phases

Phase 0: Fix Structure Violations (CRITICAL)

Fixes PRDs that are not in nested folders. This phase runs FIRST because all subsequent phases depend on correct folder structure.

Nested Folder Rule: ALL PRDs MUST be in nested folders regardless of document size.

Required Structure:

Fix Actions:

Structure Fix Workflow:

def fix_prd_structure(prd_path: str) -> list[Fix]:
    """Fix PRD structure violations."""
    fixes = []

    filename = os.path.basename(prd_path)
    parent_folder = os.path.dirname(prd_path)

    # Extract PRD ID and slug from filename
    match = re.match(r'PRD-(\d+)_([^/]+)\.md', filename)
    if not match:
        return []  # Cannot auto-fix invalid filename

    prd_id = match.group(1)
    slug = match.group(2)
    expected_folder = f"PRD-{prd_id}_{slug}"

    # Check if already in nested folder
    if os.path.basename(parent_folder) != expected_folder:
        # Create nested folder
        new_folder = os.path.join(os.path.dirname(parent_folder), expected_folder)
        os.makedirs(new_folder, exist_ok=True)

        # Move file
        new_path = os.path.join(new_folder, filename)
        shutil.move(prd_path, new_path)
        fixes.append(f"Moved {prd_path} to {new_path}")

        # Update BRD links in moved file
        content = Path(new_path).read_text()
        updated_content = content.replace('../01_BRD/', '../../01_BRD/')
        updated_content = updated_content.replace('../00_REF/', '../../00_REF/')
        Path(new_path).write_text(updated_content)
        fixes.append(f"Updated relative links for nested folder structure")

    return fixes

Link Path Updates After Move:

Phase 1: Create Missing Files

Creates files that are referenced but don't exist.

Scope:

PRD Glossary Template:

---
title: "PRD-00: Product Glossary"
tags:
  - prd
  - glossary
  - reference
custom_fields:
  document_type: glossary
  artifact_type: PRD-REFERENCE
  layer: 2
---

# PRD-00: Product Glossary

Common terminology used across all Product Requirements Documents.

## Product Terms

| Term | Definition | Context |
|------|------------|---------|
| Feature | Discrete unit of product functionality | Scope definition |
| User Story | User-centric requirement format | Requirements |
| Acceptance Criteria | Conditions for feature completion | Validation |

## Technical Terms

| Term | Definition | Context |
|------|------------|---------|
| API | Application Programming Interface | Integration |
| UI | User Interface | Frontend |
| UX | User Experience | Design |

## Domain Terms

<!-- Add project-specific terminology below -->

| Term | Definition | Context |
|------|------------|---------|
| [Term] | [Definition] | [Where used] |

Feature Placeholder Template:

---
title: "Feature Specification: [Feature Name]"
tags:
  - prd
  - feature-spec
  - reference
custom_fields:
  document_type: feature-spec
  status: placeholder
  created_by: doc-prd-fixer
---

# Feature Specification: [Feature Name]

> **Status**: Placeholder - Requires completion

## 1. Feature Overview

[TODO: Document feature overview]

## 2. User Stories

| Story ID | As a... | I want to... | So that... |
|----------|---------|--------------|------------|
| US-XX-01 | [Role] | [Action] | [Benefit] |

## 3. Acceptance Criteria

[TODO: Document acceptance criteria]

## 4. Dependencies

[TODO: Document feature dependencies]

---

*Created by doc-prd-fixer as placeholder. Complete this document to resolve broken link issues.*

Phase 2: Fix Broken Links

Updates links to point to correct locations.

Fix Actions:

Path Resolution Logic:

def fix_link_path(prd_location: str, target_path: str) -> str:
    """Calculate correct relative path based on PRD location."""

    # ALL PRDs are in nested folders (mandatory rule):
    # Monolithic PRD: docs/02_PRD/PRD-01_slug/PRD-01_slug.md
    # Sectioned PRD: docs/02_PRD/PRD-01_slug/PRD-01.3_section.md

    # Both types need to go up two levels to reach 01_BRD or 00_REF
    return "../../" + calculate_relative_path(prd_location, target_path)

BRD Link Fix (All PRDs in nested folders):

Phase 3: Fix Element IDs (BLOCKING)

Converts invalid element IDs to correct format. Element code violations are now BLOCKING - the PRD audit will fail until these are fixed.

Conversion Rules:

Type Code Mapping (PRD-specific valid codes: 01-10, 22, 24, 32):

Section-Element Type Code Mapping (BLOCKING Enforcement):

Elements MUST use the correct type code based on their section:

Fix Action for Section Mismatch (PRD-E022):

When PRD.NN.05.* IDs appear in Section 5 (Success Metrics), convert to PRD.NN.08.*:

# Section 5 metrics must use type code 08, not 05
old_id = "PRD.01.0501"  # Wrong: uses section number as type code
new_id = "PRD.01.0801"  # Correct: uses Metric type code 08

Regex Patterns:

# Find element IDs with invalid type codes for PRD
invalid_prd_type_25 = r'PRD\.(\d{2})\.25\.(\d{2})'
replacement_25 = r'PRD.\1.01.\2'

invalid_prd_type_33 = r'PRD\.(\d{2})\.33\.(\d{2})'
replacement_33 = r'PRD.\1.22.\2'

# Find legacy patterns
legacy_fr = r'###\s+FR-(\d+):'
legacy_us = r'###\s+US-(\d+):'
legacy_ac = r'###\s+AC-(\d+):'

Phase 4: Fix Content Issues

Addresses placeholders and incomplete content.

Fix Actions:

Auto-Replacements:

replacements = {
    'YYYY-MM-DDTHH:MM:SS': datetime.now().strftime('%Y-%m-%dT%H:%M:%S'),
    'YYYY-MM-DD': datetime.now().strftime('%Y-%m-%d'),
    'MM/DD/YYYY': datetime.now().strftime('%m/%d/%Y'),
    '[Current date]': datetime.now().strftime('%Y-%m-%dT%H:%M:%S'),
    '[Product Name]': extract_product_name_from_metadata(),
}

Diagram Fix Safety Rule:

• For REV-DC* issues, apply minimal edits only to diagram declaration blocks and Mermaid sequence sections.
• Do not modify non-diagram requirement content while resolving diagram contract findings.

Phase 5: Update References

Ensures traceability and cross-references are correct.

Fix Actions:

Traceability Format:

<!-- Traceability to BRD -->
@trace: BRD-01.22.01 -> PRD-01.22.01

<!-- Reference to upstream -->
@ref: [BRD-01](../../../ai_dev_ssd_flow/PROJECT/fixtures/budget_alert/BRD-01.md)

Phase 6: Handle Upstream Drift (Auto-Merge)

Automatically merges upstream BRD changes into the PRD document based on change percentage thresholds.

Drift Detection Workflow:

flowchart TD
    A[Detect Upstream BRD Changes] --> B[Calculate Change %]
    B --> C{Change Analysis}

    C -->|< 5%| D[TIER 1: Auto-Merge]
    C -->|5-15%| E[TIER 2: Auto-Merge + Detailed Log]
    C -->|> 15%| F[TIER 3: Archive + Regenerate]

    D --> D1[Add new requirements]
    D1 --> D2[Update thresholds]
    D2 --> D3[Update references]
    D3 --> D4[Increment patch: 1.0->1.0.1]

    E --> E1[Add new requirements]
    E1 --> E2[Update thresholds]
    E2 --> E3[Update references]
    E3 --> E4[Generate detailed changelog]
    E4 --> E5[Increment minor: 1.0->1.1]

    F --> F1[Mark current as ARCHIVED]
    F1 --> F2[Update status in frontmatter]
    F2 --> F3[Trigger regeneration via autopilot]
    F3 --> F4[Increment major: 1.x->2.0]

    D4 --> G[Update Drift Cache]
    E5 --> G
    F4 --> G
    G --> H[Add to Fix Report]

6.0.1 Hash Validation Fixes

FIX-H001: Invalid Hash Placeholder

Trigger: Hash contains placeholder instead of SHA-256

Fix:

sha256sum <upstream_file_path> | cut -d' ' -f1

Update cache with: sha256:<64_hex_output>

FIX-H002: Missing Hash Prefix

Trigger: 64 hex chars but missing sha256: prefix

Fix: Prepend sha256: to value

FIX-H003: Upstream File Not Found

Trigger: Cannot compute hash (file missing)

Fix: Set drift_detected: true, add to manual review

6.1 Change Percentage Calculation

def calculate_change_percentage(upstream_old: str, upstream_new: str) -> dict:
    """
    Calculate change percentage between upstream BRD versions.

    Returns:
        {
            'total_change_pct': float,      # Overall change percentage
            'additions_pct': float,          # New content added
            'modifications_pct': float,      # Existing content modified
            'deletions_pct': float,          # Content removed (tracked, not applied)
            'change_type': str               # 'minor' | 'moderate' | 'major'
        }
    """
    old_lines = upstream_old.strip().split('\n')
    new_lines = upstream_new.strip().split('\n')

    # Use difflib for precise change detection
    import difflib
    diff = difflib.unified_diff(old_lines, new_lines)

    additions = sum(1 for line in diff if line.startswith('+') and not line.startswith('+++'))
    deletions = sum(1 for line in diff if line.startswith('-') and not line.startswith('---'))

    total_lines = max(len(old_lines), len(new_lines))
    total_change_pct = ((additions + deletions) / total_lines) * 100 if total_lines > 0 else 0

    return {
        'total_change_pct': round(total_change_pct, 2),
        'additions_pct': round((additions / total_lines) * 100, 2) if total_lines > 0 else 0,
        'modifications_pct': round((min(additions, deletions) / total_lines) * 100, 2) if total_lines > 0 else 0,
        'deletions_pct': round((deletions / total_lines) * 100, 2) if total_lines > 0 else 0,
        'change_type': 'minor' if total_change_pct < 5 else 'moderate' if total_change_pct < 15 else 'major'
    }

6.2 Tier 1: Auto-Merge (< 5% Change)

Trigger: Total change percentage < 5%

Actions:

ID Generation for New Requirements:

def generate_next_id(doc_type: str, doc_num: str, element_type: str, existing_ids: list) -> str:
    """
    Generate next sequential ID for new requirement.

    Args:
        doc_type: 'PRD', 'BRD', etc.
        doc_num: '01', '02', etc.
        element_type: '01' (Functional), '05' (User Story), etc.
        existing_ids: List of existing IDs in document

    Returns:
        Next available ID (e.g., 'PRD.01.0113')
    """
    pattern = f"{doc_type}.{doc_num}.{element_type}."
    matching = [id for id in existing_ids if id.startswith(pattern)]

    if not matching:
        return f"{pattern}01"

    max_seq = max(int(id.split('.')[-1]) for id in matching)
    return f"{pattern}{str(max_seq + 1).zfill(2)}"

ID Pattern for PRD: PRD.NN.xxxx where:

• NN = Document number (01, 02, etc.)
• TT = Type code (01=Functional, 05=User Story, 06=Acceptance Criterion, etc.)
• SS = Sequence number (01, 02, etc.)

Auto-Merge Template for New Requirements:

### {GENERATED_ID}: {Requirement Title}

**Source**: Auto-merged from {upstream_brd} ({change_date})

**Requirement**: {requirement_text}

**User Stories**:
{user_stories}

**Acceptance Criteria**:
{acceptance_criteria}

**Priority**: {priority}

<!-- AUTO-MERGED: {timestamp} from {upstream_brd}#{section} -->

Version Update:

• Increment patch version: 1.0 -> 1.0.1
• Update last_updated in frontmatter
• Add changelog entry

6.3 Tier 2: Auto-Merge with Detailed Log (5-15% Change)

Trigger: Total change percentage between 5% and 15%

Actions: Same as Tier 1, plus:

Changelog Entry Format:

## Changelog

### Version 1.1 (2026-02-10T16:00:00)

**Upstream Sync**: Auto-merged 8.5% changes from upstream BRD documents

| Change | Source | Section | Description |
|--------|--------|---------|-------------|
| Added | BRD-01.1_core.md | 3.5 | New passkey authentication requirement |
| Updated | BRD-01.2_requirements.md | 4.2 | Session timeout changed 30->45 min |
| Added | BRD-01.3_quality_ops.md | 7.2 | New performance requirement |

**New Requirements Added**:
- PRD.01.0113: Passkey Authentication Support
- PRD.01.0114: WebAuthn Fallback Mechanism

**Thresholds Updated**:
- PRD.01.0205: session_idle_timeout: 30->45 min

**Impact**: EARS-01, BDD-01, ADR-01 may require review

Version Update:

• Increment minor version: 1.0 -> 1.1
• Update last_updated in frontmatter
• Add detailed changelog entry

6.4 Tier 3: Archive and Regenerate (> 15% Change)

Trigger: Total change percentage > 15%

Actions:

Archive Frontmatter Update:

---
title: "PRD-01: F1 Identity & Access Management"
custom_fields:
  version: "1.2"
  status: "archived"                    # Changed from 'current'
  archived_date: "2026-02-10T16:00:00"
  archived_reason: "upstream_drift_major"
  superseded_by: "PRD-01_v2.0"
  upstream_change_pct: 18.5
---

Archive File Naming:

docs/02_PRD/PRD-01_f1_iam/
├── PRD-01.md              # Current (v2.0)
├── PRD-01.1_core.md               # Current (v2.0)
├── .archive/
│   ├── v1.2/
│   │   ├── PRD-01.md      # Archived v1.2
│   │   ├── PRD-01.1_core.md
│   │   └── ARCHIVE_MANIFEST.md    # Archive metadata

ARCHIVE_MANIFEST.md:

# Archive Manifest: PRD-01 v1.2

| Field | Value |
|-------|-------|
| Archived Version | 1.2 |
| Archived Date | 2026-02-10T16:00:00 |
| Reason | Upstream drift > 15% (18.5%) |
| Superseded By | v2.0 |
| Upstream Changes | BRD-01 (major revision) |

## Change Summary

| Upstream Document | Change % | Key Changes |
|-------------------|----------|-------------|
| BRD-01.1_core.md | 18.5% | New auth methods, revised security model |

## Downstream Impact

Documents requiring update after regeneration:
- EARS-01 (derived from PRD-01)
- BDD-01 (test scenarios)
- ADR-01 (architecture decisions)

No Deletion Policy:

• Upstream content marked as deleted is NOT removed from document
• Instead, marked with [DEPRECATED] status:

### PRD.01.0105: Legacy Authentication Method [DEPRECATED]

> **Status**: DEPRECATED (upstream removed 2026-02-10T16:00:00)
> **Reason**: Replaced by PRD.01.0113 (Passkey Authentication)
> **Action**: Retain for traceability; do not implement

**Original Requirement**: {original_text}

6.5 Drift Cache Update

After processing drift, update .drift_cache.json:

{
  "document_version": "1.1",
  "last_synced": "2026-02-10T16:00:00",
  "sync_status": "auto-merged",
  "upstream_state": {
    "../01_BRD/BRD-01.1_core.md": {
      "hash": "sha256:a1b2c3d4e5f6...",
      "version": "2.3",
      "last_modified": "2026-02-10T15:30:00",
      "change_pct": 4.2,
      "sync_action": "tier1_auto_merge"
    },
    "../01_BRD/BRD-01.2_requirements.md": {
      "hash": "sha256:g7h8i9j0k1l2...",
      "version": "1.5",
      "last_modified": "2026-02-10T14:00:00",
      "change_pct": 8.7,
      "sync_action": "tier2_auto_merge_detailed"
    }
  },
  "merge_history": [
    {
      "date": "2026-02-10T16:00:00",
      "tier": 1,
      "change_pct": 4.2,
      "items_added": 1,
      "items_updated": 2,
      "version_before": "1.0",
      "version_after": "1.0.1"
    }
  ],
  "deprecated_items": [
    {
      "id": "PRD.01.0105",
      "deprecated_date": "2026-02-10T16:00:00",
      "reason": "Upstream removal",
      "replaced_by": "PRD.01.0113"
    }
  ]
}

6.6 Fix Report: Drift Section

Drift Summary in Fix Report:

## Phase 6: Upstream Drift Resolution

### Drift Analysis Summary

| Upstream Document | Change % | Tier | Action Taken |
|-------------------|----------|------|--------------|
| BRD-01.1_core.md | 4.2% | 1 | Auto-merged |
| BRD-01.2_requirements.md | 8.7% | 2 | Auto-merged (detailed) |
| BRD-01.3_quality_ops.md | 18.5% | 3 | Archived + Regenerated |

### Tier 1 Auto-Merges (< 5%)

| ID | Type | Source | Description |
|----|------|--------|-------------|
| PRD.01.0113 | Added | BRD-01.1:3.5 | Passkey authentication support |
| PRD.01.0205 | Updated | BRD-01.1:4.2 | Session timeout 30->45 min |

### Tier 2 Auto-Merges (5-15%)

| ID | Type | Source | Description |
|----|------|--------|-------------|
| PRD.01.0114 | Added | BRD-01.2:5.3 | WebAuthn fallback mechanism |
| PRD.01.0704 | Added | BRD-01.2:7.2 | New risk: credential phishing |

### Tier 3 Archives (> 15%)

| Document | Previous Version | New Version | Reason |
|----------|------------------|-------------|--------|
| PRD-01.2_requirements.md | 1.2 | 2.0 | 18.5% upstream change |

**Archive Location**: `docs/02_PRD/PRD-01_f1_iam/.archive/v1.2/`

### Deprecated Items (No Deletion)

| ID | Deprecated Date | Reason | Replaced By |
|----|-----------------|--------|-------------|
| PRD.01.0105 | 2026-02-10T16:00:00 | Upstream removed | PRD.01.0113 |

### Version Changes

| File | Before | After | Change Type |
|------|--------|-------|-------------|
| PRD-01.1_core.md | 1.0 | 1.0.1 | Patch (Tier 1) |
| PRD-01.2_requirements.md | 1.0 | 1.1 | Minor (Tier 2) |
| PRD-01.3_features.md | 1.2 | 2.0 | Major (Tier 3) |

Command Usage

Basic Usage

# Fix PRD based on latest review
/doc-prd-fixer PRD-01

# Fix with explicit review report (legacy)
/doc-prd-fixer PRD-01 --review-report PRD-01.R_review_report_v001.md

# Fix with explicit audit report (preferred)
/doc-prd-fixer PRD-01 --review-report PRD-01.A_audit_report_v001.md

# Fix and re-run review
/doc-prd-fixer PRD-01 --revalidate

# Fix with iteration limit
/doc-prd-fixer PRD-01 --revalidate --max-iterations 3

Options

Fix Types

Output Artifacts

Fix Report

Nested Folder Rule: ALL PRDs use nested folders (PRD-NN_{slug}/) regardless of size. Fix reports are stored alongside the PRD document in the nested folder.

File Naming: PRD-NN.F_fix_report_vNNN.md

Location: Inside the PRD nested folder: docs/02_PRD/PRD-NN_{slug}/

Structure:

---
title: "PRD-NN.F: Fix Report v001"
tags:
  - prd
  - fix-report
  - quality-assurance
custom_fields:
  document_type: fix-report
  artifact_type: PRD-FIX
  layer: 2
  parent_doc: PRD-NN
  source_review: PRD-NN.R_review_report_v001.md
  fix_date: "YYYY-MM-DDTHH:MM:SS"
  fix_tool: doc-prd-fixer
  fix_version: "2.0"
---

# PRD-NN Fix Report v001

## Summary

| Metric | Value |
|--------|-------|
| Source Review | PRD-NN.R_review_report_v001.md |
| Issues in Review | 12 |
| Issues Fixed | 10 |
| Issues Remaining | 2 (manual review required) |
| Files Created | 2 |
| Files Modified | 4 |

## Files Created

| File | Type | Location |
|------|------|----------|
| PRD-00_GLOSSARY.md | Product Glossary | docs/02_PRD/ |
| PRD-01_APPENDIX_A.md | Appendix Placeholder | docs/02_PRD/ |

## Fixes Applied

| # | Issue Code | Issue | Fix Applied | File |
|---|------------|-------|-------------|------|
| 1 | REV-L001 | Broken glossary link | Created PRD-00_GLOSSARY.md | PRD-01.3_features.md |
| 2 | REV-L004 | Broken BRD reference | Updated path to ../01_BRD/BRD-01.md | PRD-01.1_core.md |
| 3 | REV-N004 | Element type 25 invalid | Converted to type 01 | PRD-01.1_core.md |
| 4 | REV-L003 | Absolute path used | Converted to relative | PRD-01.2_requirements.md |

## Issues Requiring Manual Review

| # | Issue Code | Issue | Location | Reason |
|---|------------|-------|----------|--------|
| 1 | REV-P001 | [TODO] placeholder | PRD-01.2:L45 | Product decision needed |
| 2 | REV-P006 | Missing user story format | PRD-01.2:L120 | Story refinement required |

## Validation After Fix

| Metric | Before | After | Delta |
|--------|--------|-------|-------|
| Review Score | 92 | 97 | +5 |
| Errors | 2 | 0 | -2 |
| Warnings | 4 | 1 | -3 |

## Next Steps

1. Complete PRD-01_APPENDIX_A.md placeholder
2. Address remaining [TODO] placeholders
3. Run `/doc-prd-reviewer PRD-01` to verify fixes

Integration with Autopilot

This skill is invoked by doc-prd-autopilot in the Review -> Fix cycle:

flowchart LR
    subgraph Phase5["Phase 5: Review & Fix Cycle"]
        A[doc-prd-reviewer] --> B{Score >= 90?}
        B -->|No| C[doc-prd-fixer]
        C --> D{Iteration < Max?}
        D -->|Yes| A
        D -->|No| E[Flag for Manual Review]
        B -->|Yes| F[PASS]
    end

Autopilot Integration Points:

Error Handling

Recovery Actions

Backup Strategy

Before applying any fixes:

1. Create backup in tmp/backup/PRD-NN_YYYYMMDD_HHMMSS/
2. Copy all PRD files to backup location
3. Apply fixes to original files
4. If error during fix, restore from backup

Related Skills

Version History

Implementation Plan Consistency (IPLAN-004)

• Treat plan-derived outputs as valid source mode and verify intent preservation from implementation plan scope/objectives.
• Validate upstream autopilot precedence assumption: --iplan > --ref > --prompt.
• Flag objective/scope conflicts between plan context and artifact output as blocking issues requiring clarification.
• Do not introduce legacy fallback paths such as docs-v2.0/00_REF.