feat: Add role specifications for 三省六部 architecture

- Introduced role specifications for 尚书省 (shangshu), 刑部 (xingbu), and 中书省 (zhongshu) to facilitate task management and execution flow. - Implemented quality gates for each phase of the process to ensure compliance and quality assurance. - Established a coordinator role to manage the overall workflow and task distribution among the departments. - Created a team configuration file to define roles, responsibilities, and routing rules for task execution. - Added localization support for DeepWiki in both English and Chinese, enhancing accessibility for users.
2026-03-07 16:41:06 +08:00 · 2026-03-06 11:26:27 +08:00
parent 56c06ecf3d
commit 33cc451b61
46 changed files with 3050 additions and 1832 deletions
--- a/.claude/skills/skill-simplify/SKILL.md
+++ b/.claude/skills/skill-simplify/SKILL.md
@@ -0,0 +1,56 @@
+---
+name: skill-simplify
+description: SKILL.md simplification with functional integrity verification. Analyze redundancy, optimize content, check no functionality lost. Triggers on "simplify skill", "optimize skill", "skill-simplify".
+allowed-tools: AskUserQuestion, Read, Write, Edit, Bash, Glob, Grep
+---
+
+# Skill Simplify
+
+Three-phase pipeline: analyze functional inventory, apply optimization rules, verify integrity.
+
+**Phase Reference Documents** (read on-demand):
+
+| Phase | Document | Purpose |
+|-------|----------|---------|
+| 1 | [phases/01-analysis.md](phases/01-analysis.md) | Extract functional inventory, identify redundancy, validate pseudo-code format |
+| 2 | [phases/02-optimize.md](phases/02-optimize.md) | Apply simplification rules, fix format issues |
+| 3 | [phases/03-check.md](phases/03-check.md) | Verify functional integrity, validate format |
+
+## Input Processing
+
+```javascript
+const targetPath = input.trim()
+const targetFile = targetPath.endsWith('.md') ? targetPath : `${targetPath}/SKILL.md`
+const originalContent = Read(targetFile)
+const originalLineCount = originalContent.split('\n').length
+```
+
+## TodoWrite Pattern
+
+```javascript
+TodoWrite({ todos: [
+  { content: `Phase 1: Analyzing ${targetFile}`, status: "in_progress", activeForm: "Extracting functional inventory" },
+  { content: "Phase 2: Optimize", status: "pending" },
+  { content: "Phase 3: Integrity Check", status: "pending" }
+]})
+```
+
+## Core Rules
+
+1. **Preserve ALL functional elements**: Code blocks with logic, agent calls, data structures, routing, error handling, input/output specs
+2. **Only reduce descriptive content**: Flowcharts, verbose comments, duplicate sections, examples that repeat logic
+3. **Never summarize algorithm logic**: If-else branches, function bodies, schemas must remain verbatim
+4. **Classify code blocks**: Distinguish `functional` (logic, routing, schemas) from `descriptive` (ASCII art, examples, display templates) — only descriptive blocks may be deleted
+5. **Merge equivalent variants**: Single/multi-perspective templates differing only by a parameter → one template with variant comment
+6. **Fix format issues**: Nested backtick template literals in code fences → convert to prose; hardcoded option lists → flag for dynamic generation; workflow handoff references → ensure execution steps present
+7. **Validate pseudo-code**: Check bracket matching, variable consistency, structural completeness
+8. **Quantitative verification**: Phase 3 counts must match Phase 1 counts for functional categories; descriptive block decreases are expected
+
+## Error Handling
+
+| Error | Resolution |
+|-------|------------|
+| Target file not found | Report error, stop |
+| Check FAIL (missing functional elements) | Show delta, revert to original, report which elements lost |
+| Check WARN (descriptive decrease or merge) | Show delta with justification |
+| Format issues found | Report in check, fix in Phase 2 |
--- a/.claude/skills/skill-simplify/phases/01-analysis.md
+++ b/.claude/skills/skill-simplify/phases/01-analysis.md
@@ -0,0 +1,224 @@
+# Phase 1: Functional Analysis
+
+Read target file, extract functional inventory with code block classification, identify redundancy, validate pseudo-code format, and produce optimization plan.
+
+## Objective
+
+- Build quantitative functional inventory with code block classification (baseline for Phase 3)
+- Identify redundancy categories with specific line ranges
+- Detect pseudo-code format issues
+- Produce optimization plan with estimated line savings
+
+## Execution
+
+### Step 1.1: Read & Measure Target
+
+```javascript
+const originalContent = Read(targetFile)
+const lines = originalContent.split('\n')
+const originalLineCount = lines.length
+```
+
+### Step 1.2: Extract Functional Inventory
+
+Count and catalog every functional element. These counts are the **baseline** for Phase 3 verification.
+
+```javascript
+const inventory = {
+  // Code structures — with role classification
+  codeBlocks: [],        // { startLine, endLine, language, purpose, role: 'functional'|'descriptive' }
+  agentCalls: [],        // { line, agentType, description, mergeGroup?: string }
+  dataStructures: [],    // { line, name, type: 'object'|'array'|'schema' }
+
+  // Logic elements
+  routingBranches: [],   // { line, condition, outcomes[] }
+  errorHandlers: [],     // { line, errorType, resolution }
+  conditionalLogic: [],  // { line, condition, trueAction, falseAction }
+
+  // Interface elements
+  askUserQuestions: [],   // { line, questionCount, headers[], optionType: 'static'|'dynamic' }
+  inputModes: [],        // { line, mode, description }
+  outputArtifacts: [],   // { line, artifact, format }
+
+  // Structural elements
+  todoWriteBlocks: [],   // { line, phaseCount }
+  phaseHandoffs: [],     // { line, fromPhase, toPhase }
+  skillInvocations: [],  // { line, skillName, hasExecutionSteps: boolean }
+
+  // Reference elements
+  tables: [],            // { startLine, endLine, columns }
+  schemas: [],           // { line, schemaName, fields[] }
+
+  // Format issues
+  formatIssues: [],      // { line, type, description, severity: 'error'|'warning' }
+
+  // Totals (computed)
+  counts: {}
+}
+```
+
+**Extraction rules**:
+- **Code blocks**: Match ` ```language ... ``` ` pairs, record start/end/language/first-line-as-purpose
+- **Agent calls**: Match `Agent(`, `Task(`, `subagent_type=`, record type and prompt summary
+- **Data structures**: Match `const xxx = {`, `const xxx = [`, JSON schema objects
+- **Routing branches**: Match `if/else`, `switch/case`, ternary `? :` with meaningful branching
+- **Error handlers**: Match `catch`, error table rows `| Error |`, fallback patterns
+- **AskUserQuestion**: Match `AskUserQuestion({`, count questions array length
+- **Input modes**: Match `Mode 1/2/3`, `--flag`, argument parsing
+- **Output artifacts**: Match `Write(`, `Output:`, file path patterns in comments
+- **TodoWrite**: Match `TodoWrite({`, count todo items
+- **Phase handoffs**: Match `Read("phases/`, `Skill(`, `proceed_to_next_phase`
+- **Tables**: Match `| header |` markdown table blocks
+- **Schemas**: Match schema references, JSON structure definitions
+
+### Step 1.2.1: Code Block Role Classification
+
+For each code block, determine its role:
+
+| Role | Criteria | Examples |
+|------|----------|---------|
+| `functional` | Contains algorithm logic, routing branches, conditional code, agent calls, schema definitions, data processing, AskUserQuestion, Skill invocations | `if/else`, `Agent({...})`, `const schema = {...}`, `Bash({...})` |
+| `descriptive` | Contains ASCII art, usage examples, display templates, illustrative good/bad comparisons, folder structure trees | `┌───┐`, `# Example usage`, `❌ Bad / ✅ Good`, `├── file.ts` |
+
+**Classification rules**:
+- If block contains ANY of: `Agent(`, `Bash(`, `AskUserQuestion(`, `if (`, `switch`, `Skill(`, `Write(`, `Read(`, `TodoWrite(` → `functional`
+- If block language is `bash` and content is only example invocations (no logic) → `descriptive`
+- If block has no language tag and contains only ASCII box-drawing characters → `descriptive`
+- If block is labeled as "Example" in surrounding markdown heading → `descriptive`
+- **Default**: `functional` (conservative)
+
+### Step 1.2.2: Pseudo-Code Format Validation
+
+Scan all `functional` code blocks for format issues:
+
+| Check | Detection | Severity |
+|-------|-----------|----------|
+| **Nested backticks** | Template literal `` ` `` inside ` ```javascript ``` ` code fence | warning |
+| **Unclosed brackets** | Unmatched `{`, `(`, `[` in code block | error |
+| **Undefined references** | `${variable}` where variable is never declared in the block or prior blocks | warning |
+| **Inconsistent indentation** | Mixed tabs/spaces or inconsistent nesting depth | warning |
+| **Dead code patterns** | Commented-out code blocks (`// if (`, `/* ... */` spanning 5+ lines) | warning |
+| **Missing return/output** | Function-like block with no return, Write, or console.log | warning |
+
+```javascript
+inventory.formatIssues = validatePseudoCode(inventory.codeBlocks.filter(b => b.role === 'functional'))
+```
+
+### Step 1.2.3: Compute Totals
+
+```javascript
+inventory.counts = {
+  codeBlocks: inventory.codeBlocks.length,
+  functionalCodeBlocks: inventory.codeBlocks.filter(b => b.role === 'functional').length,
+  descriptiveCodeBlocks: inventory.codeBlocks.filter(b => b.role === 'descriptive').length,
+  agentCalls: inventory.agentCalls.length,
+  dataStructures: inventory.dataStructures.length,
+  routingBranches: inventory.routingBranches.length,
+  errorHandlers: inventory.errorHandlers.length,
+  conditionalLogic: inventory.conditionalLogic.length,
+  askUserQuestions: inventory.askUserQuestions.length,
+  inputModes: inventory.inputModes.length,
+  outputArtifacts: inventory.outputArtifacts.length,
+  todoWriteBlocks: inventory.todoWriteBlocks.length,
+  phaseHandoffs: inventory.phaseHandoffs.length,
+  skillInvocations: inventory.skillInvocations.length,
+  tables: inventory.tables.length,
+  schemas: inventory.schemas.length,
+  formatIssues: inventory.formatIssues.length
+}
+```
+
+### Step 1.3: Identify Redundancy Categories
+
+Scan for each category, record specific line ranges:
+
+```javascript
+const redundancyMap = {
+  deletable: [],       // { category, startLine, endLine, reason, estimatedSave }
+  simplifiable: [],    // { category, startLine, endLine, strategy, estimatedSave }
+  mergeable: [],       // { items: [{startLine, endLine}], mergeStrategy, estimatedSave }
+  formatFixes: [],     // { line, type, fix }
+  languageUnify: []    // { line, currentLang, targetLang }
+}
+```
+
+**Deletable** (remove entirely, no functional loss):
+
+| Pattern | Detection |
+|---------|-----------|
+| Duplicate Overview | `## Overview` that restates frontmatter description |
+| ASCII flowchart | Flowchart that duplicates Phase Summary table or implementation structure |
+| "When to use" section | Usage guidance not needed for execution |
+| Best Practices section | Advisory content duplicating Core Rules |
+| Duplicate examples | Code examples that repeat logic shown elsewhere |
+| Folder structure duplicate | ASCII tree repeating Output Artifacts table |
+| "Next Phase" paragraphs | Prose between phases when TodoWrite handles flow |
+| Descriptive code blocks | Code blocks classified as `descriptive` whose content is covered by surrounding prose or tables |
+
+**Simplifiable** (compress, preserve meaning):
+
+| Pattern | Strategy |
+|---------|----------|
+| Verbose comments in code blocks | Reduce to single-line; keep only non-obvious logic comments |
+| Multi-line console.log | Compress to single template literal |
+| Wordy section intros | Remove "In this phase, we will..." preamble |
+| Exploration prompt bloat | Trim to essential instructions, remove generic advice |
+| Display-format code blocks | Convert code blocks that only define output format (console.log with template) to prose description |
+
+**Mergeable** (combine related structures):
+
+| Pattern | Strategy |
+|---------|----------|
+| Multiple similar AskUserQuestion calls | Extract shared function with mode parameter |
+| Repeated Option routing | Unify into single dispatch |
+| Sequential single-line operations | Combine into one code block |
+| TodoWrite full blocks x N | Template once + delta comments |
+| Duplicate error handling tables | Merge into single table |
+| Equivalent template variants | Single/multi-perspective templates → one template with variant comment |
+| Multiple output artifact tables | Merge into single combined table |
+
+**Format fixes** (pseudo-code quality):
+
+| Pattern | Fix |
+|---------|-----|
+| Nested backtick template literals | Convert surrounding code block to prose description, or use 4-backtick fence |
+| Hardcoded option lists | Add comment: `// Generate dynamically from {context source}` |
+| Workflow handoff without execution steps | Add execution steps referencing the target command's actual interface |
+| Unclosed brackets | Fix bracket matching |
+
+**Language unification**:
+- Detect mixed Chinese/English in functional comments
+- Recommend consistent language (match majority)
+
+### Step 1.4: Build Optimization Plan
+
+```javascript
+const optimizationPlan = {
+  targetFile,
+  originalLineCount,
+  estimatedReduction: redundancyMap.deletable.reduce((s, d) => s + d.estimatedSave, 0)
+    + redundancyMap.simplifiable.reduce((s, d) => s + d.estimatedSave, 0)
+    + redundancyMap.mergeable.reduce((s, d) => s + d.estimatedSave, 0),
+  categories: {
+    deletable: { count: redundancyMap.deletable.length, totalLines: '...' },
+    simplifiable: { count: redundancyMap.simplifiable.length, totalLines: '...' },
+    mergeable: { count: redundancyMap.mergeable.length, totalLines: '...' },
+    formatFixes: { count: redundancyMap.formatFixes.length },
+    languageUnify: { count: redundancyMap.languageUnify.length }
+  },
+  // Ordered: delete → merge → simplify → format
+  operations: [
+    ...redundancyMap.deletable.map(d => ({ type: 'delete', ...d, priority: 1 })),
+    ...redundancyMap.mergeable.map(m => ({ type: 'merge', ...m, priority: 2 })),
+    ...redundancyMap.simplifiable.map(s => ({ type: 'simplify', ...s, priority: 3 })),
+    ...redundancyMap.formatFixes.map(f => ({ type: 'format', ...f, priority: 4 }))
+  ]
+}
+```
+
+Display plan summary: category counts, estimated reduction percentage, sections NOT changed (functional core).
+
+## Output
+
+- **Variable**: `analysisResult = { inventory, redundancyMap, optimizationPlan, originalContent, originalLineCount }`
+- **TodoWrite**: Mark Phase 1 completed, Phase 2 in_progress
--- a/.claude/skills/skill-simplify/phases/02-optimize.md
+++ b/.claude/skills/skill-simplify/phases/02-optimize.md
@@ -0,0 +1,107 @@
+# Phase 2: Optimize
+
+Apply simplification rules from analysisResult to produce optimized content. Write result to disk.
+
+## Objective
+
+- Execute all optimization operations in priority order (delete → merge → simplify → format)
+- Preserve every functional element identified in Phase 1 inventory
+- Fix pseudo-code format issues
+- Write optimized content back to target file
+
+## Execution
+
+### Step 2.1: Apply Operations in Order
+
+Process `analysisResult.optimizationPlan.operations` sorted by priority:
+
+**Priority 1 — Delete** (safest, highest impact):
+
+| Target Pattern | Action |
+|----------------|--------|
+| Duplicate Overview section | Remove `## Overview` if it restates frontmatter `description` |
+| ASCII flowchart | Remove if Phase Summary table or implementation structure covers same info |
+| "When to use" / "Use Cases" section | Remove entirely |
+| Best Practices section | Remove if content duplicates Core Rules |
+| Duplicate folder structure | Remove ASCII tree if Output Artifacts table covers same info |
+| Redundant "Next Phase" prose | Remove when TodoWrite handles flow |
+| Standalone example sections | Remove if logic already demonstrated inline |
+| Descriptive code blocks | Remove if content covered by surrounding prose or tables |
+
+**Priority 2 — Merge** (structural optimization):
+
+| Target Pattern | Action |
+|----------------|--------|
+| Multiple similar AskUserQuestion blocks | Extract shared function with mode parameter |
+| Repeated Option A/B/C routing | Unify into single dispatch |
+| Sequential single-line bash commands | Combine into single code block |
+| TodoWrite full blocks x N | Template ONCE, subsequent as one-line comment |
+| Duplicate error handling across sections | Merge into single `## Error Handling` table |
+| Equivalent template variants | Single/multi templates → one template with `// For multi: add Perspective` comment |
+| Multiple output artifact tables | Merge into single combined table with Phase column |
+
+**Priority 3 — Simplify** (compress descriptive content):
+
+| Target Pattern | Action |
+|----------------|--------|
+| Verbose inline comments | Reduce to single-line; remove obvious restatements |
+| Display-format code blocks | Convert `console.log` with template literal to prose describing output format |
+| Wordy section introductions | Remove preamble sentences |
+| Exploration/agent prompt padding | Remove generic advice |
+| Success Criteria lists > 7 items | Trim to essential 5-7, remove obvious/generic |
+
+**Priority 4 — Format fixes** (pseudo-code quality):
+
+| Target Pattern | Action |
+|----------------|--------|
+| Nested backtick template literals | Convert code block to prose description, or use 4-backtick fence |
+| Hardcoded option lists | Replace with dynamic generation: describe source of options + generation logic |
+| Workflow handoff without execution steps | Add concrete steps referencing target command's interface (e.g., pipe to `ccw issue create`) |
+| Unclosed brackets | Fix bracket matching |
+| Undefined variable references | Add declaration or link to source |
+
+### Step 2.2: Language Unification (if applicable)
+
+```javascript
+if (analysisResult.redundancyMap.languageUnify.length > 0) {
+  // Detect majority language, unify non-functional text
+  // DO NOT change: variable names, function names, schema fields, error messages in code
+}
+```
+
+### Step 2.3: Write Optimized Content
+
+```javascript
+Write(targetFile, optimizedContent)
+const optimizedLineCount = optimizedContent.split('\n').length
+const reduction = originalLineCount - optimizedLineCount
+const reductionPct = Math.round(reduction / originalLineCount * 100)
+```
+
+### Step 2.4: Preserve Optimization Record
+
+```javascript
+const optimizationRecord = {
+  deletedSections: [],    // section names removed
+  mergedGroups: [],       // { from: [sections], to: description }
+  simplifiedAreas: [],    // { section, strategy }
+  formatFixes: [],        // { line, type, fix }
+  linesBefore: originalLineCount,
+  linesAfter: optimizedLineCount
+}
+```
+
+## Key Rules
+
+1. **Never modify functional code blocks** — only compress comments/whitespace within them
+2. **Descriptive code blocks may be deleted** if their content is covered by prose or tables
+3. **Never change function signatures, variable names, or schema fields**
+4. **Merge preserves all branches** — unified function must handle all original cases
+5. **When uncertain, keep original** — conservative approach prevents functional loss
+6. **Format fixes must not alter semantics** — only presentation changes
+
+## Output
+
+- **File**: Target file overwritten with optimized content
+- **Variable**: `optimizationRecord` (changes log for Phase 3)
+- **TodoWrite**: Mark Phase 2 completed, Phase 3 in_progress
--- a/.claude/skills/skill-simplify/phases/03-check.md
+++ b/.claude/skills/skill-simplify/phases/03-check.md
@@ -0,0 +1,224 @@
+# Phase 3: Integrity Check
+
+Re-extract functional inventory from optimized file, compare against Phase 1 baseline, validate pseudo-code format. Report PASS/FAIL with detailed delta.
+
+## Objective
+
+- Re-run the same inventory extraction on optimized content
+- Compare counts using role-aware classification (functional vs descriptive)
+- Validate pseudo-code format issues are resolved
+- Report check result with actionable details
+- Revert if critical functional elements are missing
+
+## Execution
+
+### Step 3.1: Re-Extract Inventory from Optimized File
+
+```javascript
+const optimizedContent = Read(targetFile)
+const optimizedLineCount = optimizedContent.split('\n').length
+
+// Use SAME extraction logic as Phase 1 (including role classification)
+const afterInventory = extractFunctionalInventory(optimizedContent)
+```
+
+### Step 3.2: Compare Inventories (Role-Aware)
+
+```javascript
+const beforeCounts = analysisResult.inventory.counts
+const afterCounts = afterInventory.counts
+
+const delta = {}
+let hasCriticalLoss = false
+let hasWarning = false
+
+// CRITICAL: Functional elements that MUST NOT decrease
+const CRITICAL = ['functionalCodeBlocks', 'dataStructures', 'routingBranches',
+                  'errorHandlers', 'conditionalLogic', 'askUserQuestions',
+                  'inputModes', 'outputArtifacts', 'skillInvocations']
+
+// MERGE_AWARE: May decrease due to valid merge operations — verify coverage
+const MERGE_AWARE = ['agentCalls', 'codeBlocks']
+
+// EXPECTED_DECREASE: May decrease from merge/consolidation
+const EXPECTED_DECREASE = ['descriptiveCodeBlocks', 'todoWriteBlocks',
+                           'phaseHandoffs', 'tables', 'schemas']
+
+for (const [key, before] of Object.entries(beforeCounts)) {
+  const after = afterCounts[key] || 0
+  const diff = after - before
+  let category, status
+
+  if (CRITICAL.includes(key)) {
+    category = 'critical'
+    status = diff < 0 ? 'FAIL' : 'OK'
+    if (diff < 0) hasCriticalLoss = true
+  } else if (MERGE_AWARE.includes(key)) {
+    category = 'merge_aware'
+    // Decrease is WARN (needs justification), not FAIL
+    status = diff < 0 ? 'WARN' : 'OK'
+    if (diff < 0) hasWarning = true
+  } else {
+    category = 'expected'
+    status = 'OK'  // Descriptive decreases are expected
+  }
+
+  delta[key] = { before, after, diff, category, status }
+}
+```
+
+### Step 3.3: Deep Verification
+
+**For CRITICAL categories with decrease** — identify exactly what was lost:
+
+```javascript
+if (hasCriticalLoss) {
+  const lostElements = {}
+  for (const [key, d] of Object.entries(delta)) {
+    if (d.status === 'FAIL') {
+      const beforeItems = analysisResult.inventory[key]
+      const afterItems = afterInventory[key]
+      lostElements[key] = beforeItems.filter(beforeItem =>
+        !afterItems.some(afterItem => matchesElement(beforeItem, afterItem))
+      )
+    }
+  }
+}
+```
+
+**For MERGE_AWARE categories with decrease** — verify merged coverage:
+
+```javascript
+if (hasWarning) {
+  for (const [key, d] of Object.entries(delta)) {
+    if (d.category === 'merge_aware' && d.diff < 0) {
+      // Check if merged template covers all original variants
+      // e.g., single Agent template with "// For multi: add Perspective" covers both
+      const beforeItems = analysisResult.inventory[key]
+      const afterItems = afterInventory[key]
+      const unmatched = beforeItems.filter(beforeItem =>
+        !afterItems.some(afterItem => matchesElement(beforeItem, afterItem))
+      )
+      if (unmatched.length > 0) {
+        // Check if unmatched items are covered by merge comments in remaining items
+        const mergeComments = afterItems.flatMap(item => extractMergeComments(item))
+        const trulyLost = unmatched.filter(item =>
+          !mergeComments.some(comment => coversElement(comment, item))
+        )
+        if (trulyLost.length > 0) {
+          delta[key].status = 'FAIL'
+          hasCriticalLoss = true
+          delta[key].trulyLost = trulyLost
+        }
+        // else: merge-covered, WARN is correct
+      }
+    }
+  }
+}
+```
+
+### Step 3.4: Pseudo-Code Format Validation
+
+```javascript
+const afterFormatIssues = validatePseudoCode(afterInventory.codeBlocks.filter(b => b.role === 'functional'))
+const beforeFormatCount = analysisResult.inventory.formatIssues.length
+const afterFormatCount = afterFormatIssues.length
+
+const formatDelta = {
+  before: beforeFormatCount,
+  after: afterFormatCount,
+  resolved: beforeFormatCount - afterFormatCount,
+  newIssues: afterFormatIssues.filter(issue =>
+    !analysisResult.inventory.formatIssues.some(orig => orig.line === issue.line && orig.type === issue.type)
+  )
+}
+
+// New format issues introduced by optimization = FAIL
+if (formatDelta.newIssues.length > 0) {
+  hasCriticalLoss = true
+}
+```
+
+**Pseudo-code validation checks**:
+
+| Check | Detection | Action on Failure |
+|-------|-----------|-------------------|
+| Bracket matching | Count `{([` vs `})]` per code block | FAIL — fix or revert |
+| Variable consistency | `${var}` used but never declared | WARNING — note in report |
+| Structural completeness | Function body has entry but no exit (return/Write/output) | WARNING |
+| Nested backtick resolution | Backtick template literals inside code fences | WARNING if pre-existing, FAIL if newly introduced |
+| Schema field preservation | Schema fields in after match before | FAIL if fields lost |
+
+### Step 3.5: Generate Check Report
+
+```javascript
+const status = hasCriticalLoss ? 'FAIL' : (hasWarning ? 'WARN' : 'PASS')
+
+const checkReport = {
+  status,
+  linesBefore: analysisResult.originalLineCount,
+  linesAfter: optimizedLineCount,
+  reduction: `${analysisResult.originalLineCount - optimizedLineCount} lines (-${Math.round((analysisResult.originalLineCount - optimizedLineCount) / analysisResult.originalLineCount * 100)}%)`,
+  delta,
+  formatDelta,
+  lostElements: hasCriticalLoss ? lostElements : null
+}
+
+// Display report table
+// | Category | Before | After | Delta | Status |
+// Show all categories, highlight FAIL/WARN rows
+// Show format issues summary if any
+```
+
+### Step 3.6: Act on Result
+
+```javascript
+if (status === 'FAIL') {
+  Write(targetFile, analysisResult.originalContent)
+  // Report: "Critical elements lost / new format issues introduced. Reverted."
+}
+
+if (status === 'WARN') {
+  // Report: "Decreases from merge/descriptive removal. Verify coverage."
+  // Show merge justifications for MERGE_AWARE categories
+}
+
+if (status === 'PASS') {
+  // Report: "All functional elements preserved. Optimization successful."
+}
+```
+
+## Element Matching Rules
+
+How `matchesElement()` determines if a before-element exists in after-inventory:
+
+| Element Type | Match Criteria |
+|-------------|---------------|
+| codeBlocks | Same language + first meaningful line (ignore whitespace/comments) |
+| agentCalls | Same agentType + similar prompt keywords (>60% overlap) |
+| dataStructures | Same variable name OR same field set |
+| routingBranches | Same condition expression (normalized) |
+| errorHandlers | Same error type/pattern |
+| conditionalLogic | Same condition + same outcome set |
+| askUserQuestions | Same question count + similar option labels |
+| inputModes | Same mode identifier |
+| outputArtifacts | Same file path pattern or artifact name |
+| skillInvocations | Same skill name |
+| todoWriteBlocks | Same phase names (order-independent) |
+| phaseHandoffs | Same target phase reference |
+| tables | Same column headers |
+| schemas | Same schema name or field set |
+
+**Merge coverage check** (`coversElement()`):
+- Agent calls: Merged template contains `// For multi:` or `// Multi-perspective:` comment referencing the missing variant
+- Code blocks: Merged block contains comment noting the alternative was folded in
+
+## Completion
+
+```javascript
+TodoWrite({ todos: [
+  { content: `Phase 1: Analysis [${Object.keys(analysisResult.inventory.counts).length} categories]`, status: "completed" },
+  { content: `Phase 2: Optimize [${checkReport.reduction}]`, status: "completed" },
+  { content: `Phase 3: Check [${checkReport.status}] | Format: ${formatDelta.resolved} resolved, ${formatDelta.newIssues.length} new`, status: "completed" }
+]})
+```