[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M6ESVEPBF8YT5M4MM4SGCK8S":3,"skill-same:01M6ESVEPBF8YT5M4MM4SGCK8S":125},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":26,"publishedAt":47,"linkOnly":48,"markdownContent":49,"markdownBody":50,"frontMatter":51,"refName":82,"commitSha":83,"skillDirectory":84,"forksCount":85,"repositoryHtmlUrl":86,"repositoryHomepageUrl":87,"ownerHtmlUrl":88,"ownerBlogUrl":6,"ownerDescription":89,"updatedAt":90,"install":91,"resources":104},"01M6ESVEPBF8YT5M4MM4SGCK8S","when-optimizing-prompts-use-prompt-architect",null,"Comprehensive framework for analyzing, creating, and refining prompts for AI systems using evidence-based techniques","ComeOnOliver\u002Fskillshub","ComeOnOliver","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F43215937?v=4","01M6E5JNFECJAAVFE0W7SMWWD2","Haoran Shu","01M6E5JNNDQS7AJZDYFQ8TPZCK",67,"MIT","https:\u002F\u002Fgithub.com\u002FComeOnOliver\u002Fskillshub\u002Fblob\u002Fmain\u002Fskills\u002Faiskillstore\u002Fmarketplace\u002Fdnyoussef\u002Fwhen-optimizing-prompts-use-prompt-architect\u002FSKILL.md",[18,23],{"categoryNo":19,"categoryName":20,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000008","测试与质量","SCENARIO","使用场景",{"categoryNo":24,"categoryName":25,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000023","Skill与智能体开发",[27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46],"when optimizing prompts use prompt architect","prompt architect","architect","prompt","use","prompts","Optimizing Prompts","anti-patterns","Utilities","structure","optimize","patterns","prompt-engineering","testing","refactoring","analysis","flow","analyzing","alpha","framework","2026-11-10 10:12:30",false,"---\nname: when-optimizing-prompts-use-prompt-architect\nversion: 1.0.0\ndescription: Comprehensive framework for analyzing, creating, and refining prompts for AI systems using evidence-based techniques\ncategory: utilities\ntags: [prompt-engineering, optimization, ai-systems, llm]\nagents: [researcher, coder]\ndifficulty: intermediate\nestimated_duration: 20-40min\nsuccess_criteria:\n  - Prompt structure optimized\n  - Anti-patterns eliminated\n  - Evidence-based techniques applied\n  - Effectiveness validated\nvalidation_method: A\u002FB testing\ndependencies:\n  - claude-flow@alpha\nprerequisites:\n  - Prompt to optimize OR use case defined\noutputs:\n  - Optimized prompt\n  - Analysis report\n  - A\u002FB test results\ntriggers:\n  - Poor AI response quality\n  - Inconsistent outputs\n  - New prompt needed\n  - Prompt refactoring required\n---\n\n# Prompt Architect - Evidence-Based Prompt Engineering\n\n## Overview\n\nComprehensive framework for analyzing, creating, and refining prompts for AI systems (Claude, GPT, etc.). Applies structural optimization, self-consistency patterns, and anti-pattern detection to transform prompts into highly effective versions.\n\n## When to Use This Skill\n\n- Creating new prompts for AI systems\n- Existing prompts produce poor results\n- Inconsistent AI outputs\n- Need to improve prompt clarity\n- Applying evidence-based prompt engineering\n- Optimizing agent instructions\n- Building prompt libraries\n\n## Theoretical Foundation\n\n### Evidence-Based Techniques\n\n1. **Chain-of-Thought (CoT)**: Explicit reasoning steps\n2. **Self-Consistency**: Multiple reasoning paths\n3. **ReAct**: Reasoning + Acting pattern\n4. **Program-of-Thought**: Structured logic\n5. **Plan-and-Solve**: Decomposition strategy\n6. **Role-Playing**: Persona assignment\n7. **Few-Shot Learning**: Example-based instruction\n\n### Prompt Structure Principles\n\n```\n[System Context] → [Role Definition] → [Task Description] →\n[Constraints] → [Format Specification] → [Examples] → [Quality Criteria]\n```\n\n## Phase 1: Analyze Current Prompt\n\n### Objective\nIdentify weaknesses and improvement opportunities\n\n### Agent: Researcher\n\n**Step 1.1: Structural Analysis**\n```javascript\nconst promptAnalysis = {\n  components: {\n    hasSystemContext: checkForContext(prompt),\n    hasRoleDefinition: checkForRole(prompt),\n    hasTaskDescription: checkForTask(prompt),\n    hasConstraints: checkForConstraints(prompt),\n    hasFormatSpec: checkForFormat(prompt),\n    hasExamples: checkForExamples(prompt),\n    hasQualityCriteria: checkForCriteria(prompt)\n  },\n  metrics: {\n    length: prompt.length,\n    clarity: calculateClarity(prompt),\n    specificity: calculateSpecificity(prompt),\n    completeness: calculateCompleteness(prompt)\n  },\n  antiPatterns: detectAntiPatterns(prompt)\n};\n\nawait memory.store('prompt-architect\u002Fanalysis', promptAnalysis);\n```\n\n**Step 1.2: Detect Anti-Patterns**\n```javascript\nconst antiPatterns = [\n  {\n    name: 'Vague Instructions',\n    pattern: \u002Fplease|try to|maybe|possibly\u002Fgi,\n    severity: 'HIGH',\n    fix: 'Use imperative commands: \"Analyze...\", \"Generate...\", \"Create...\"'\n  },\n  {\n    name: 'Missing Context',\n    pattern: absence of background info,\n    severity: 'HIGH',\n    fix: 'Add system context and domain information'\n  },\n  {\n    name: 'No Output Format',\n    pattern: absence of format specification,\n    severity: 'MEDIUM',\n    fix: 'Specify exact output format (JSON, markdown, etc.)'\n  },\n  {\n    name: 'Conflicting Instructions',\n    pattern: detectContradictions(prompt),\n    severity: 'HIGH',\n    fix: 'Resolve contradictions, prioritize requirements'\n  },\n  {\n    name: 'Implicit Assumptions',\n    pattern: detectImplicitAssumptions(prompt),\n    severity: 'MEDIUM',\n    fix: 'Make all assumptions explicit'\n  }\n];\n\nconst foundAntiPatterns = antiPatterns.filter(ap =>\n  ap.pattern.test ? ap.pattern.test(prompt) : ap.pattern\n);\n\nawait memory.store('prompt-architect\u002Fanti-patterns', foundAntiPatterns);\n```\n\n**Step 1.3: Identify Missing Components**\n```javascript\nconst missingComponents = [];\n\nif (!promptAnalysis.components.hasSystemContext) {\n  missingComponents.push({\n    component: 'System Context',\n    importance: 'HIGH',\n    recommendation: 'Add background info, domain knowledge, constraints'\n  });\n}\n\nif (!promptAnalysis.components.hasExamples) {\n  missingComponents.push({\n    component: 'Examples',\n    importance: 'MEDIUM',\n    recommendation: 'Add 2-3 examples showing desired behavior'\n  });\n}\n\n\u002F\u002F ... check other components\n\nawait memory.store('prompt-architect\u002Fmissing', missingComponents);\n```\n\n### Validation Criteria\n- [ ] All 7 components checked\n- [ ] Anti-patterns identified\n- [ ] Missing components listed\n- [ ] Severity assigned to issues\n\n### Hooks Integration\n```bash\nnpx claude-flow@alpha hooks pre-task \\\n  --description \"Analyze prompt structure and quality\" \\\n  --complexity \"medium\"\n\nnpx claude-flow@alpha hooks post-task \\\n  --task-id \"prompt-analysis\" \\\n  --output \"analysis-report.json\"\n```\n\n## Phase 2: Structure Optimization\n\n### Objective\nReorganize prompt for logical flow and clarity\n\n### Agent: Coder (Prompt Specialist)\n\n**Step 2.1: Apply Template Structure**\n```javascript\nconst optimizedStructure = {\n  systemContext: {\n    domain: extractDomain(prompt),\n    background: generateContextualBackground(),\n    constraints: extractOrInferConstraints(prompt)\n  },\n  roleDefinition: {\n    persona: definePersona(task),\n    expertise: listRequiredExpertise(),\n    perspective: defineWorkingPerspective()\n  },\n  taskDescription: {\n    primary: extractPrimaryTask(prompt),\n    secondary: extractSecondaryTasks(prompt),\n    scope: defineScope(),\n    outOfScope: defineWhatNotToDo()\n  },\n  constraints: {\n    required: extractRequirements(prompt),\n    forbidden: extractProhibitions(prompt),\n    optional: extractPreferences(prompt)\n  },\n  formatSpecification: {\n    outputFormat: specifyFormat(),\n    structure: defineStructure(),\n    examples: []\n  },\n  qualityCriteria: {\n    success: defineSuccessCriteria(),\n    validation: defineValidationMethod(),\n    metrics: defineMetrics()\n  }\n};\n\nawait memory.store('prompt-architect\u002Fstructure', optimizedStructure);\n```\n\n**Step 2.2: Build Optimized Prompt**\n```markdown\n# System Context\n${optimizedStructure.systemContext.background}\n\nDomain: ${optimizedStructure.systemContext.domain}\nConstraints: ${optimizedStructure.systemContext.constraints.join(', ')}\n\n# Role\nYou are ${optimizedStructure.roleDefinition.persona} with expertise in:\n${optimizedStructure.roleDefinition.expertise.map(e => `- ${e}`).join('\\n')}\n\nYour perspective: ${optimizedStructure.roleDefinition.perspective}\n\n# Task\nPrimary objective: ${optimizedStructure.taskDescription.primary}\n\n${optimizedStructure.taskDescription.secondary.length > 0 ?\n  'Secondary objectives:\\n' + optimizedStructure.taskDescription.secondary.map(t => `- ${t}`).join('\\n') : ''}\n\nScope: ${optimizedStructure.taskDescription.scope}\nOut of scope: ${optimizedStructure.taskDescription.outOfScope.join(', ')}\n\n# Constraints\nMUST: ${optimizedStructure.constraints.required.map(r => `- ${r}`).join('\\n')}\nMUST NOT: ${optimizedStructure.constraints.forbidden.map(f => `- ${f}`).join('\\n')}\nPREFER: ${optimizedStructure.constraints.optional.map(o => `- ${o}`).join('\\n')}\n\n# Output Format\n${optimizedStructure.formatSpecification.outputFormat}\n\nStructure:\n${optimizedStructure.formatSpecification.structure}\n\n# Quality Criteria\nSuccess is defined as:\n${optimizedStructure.qualityCriteria.success.map(s => `- ${s}`).join('\\n')}\n\nValidation method: ${optimizedStructure.qualityCriteria.validation}\n```\n\n**Step 2.3: Add Progressive Disclosure**\n```javascript\n\u002F\u002F For complex prompts, use hierarchical structure\nconst progressivePrompt = {\n  essential: generateEssentialInstructions(),\n  details: generateDetailedGuidance(),\n  advanced: generateAdvancedTechniques(),\n  examples: generateExamples(),\n  troubleshooting: generateTroubleshootingGuide()\n};\n\n\u002F\u002F Structure with collapsible sections\nconst enhancedPrompt = `\n${progressivePrompt.essential}\n\n\u003Cdetails>\n\u003Csummary>Detailed Guidance\u003C\u002Fsummary>\n${progressivePrompt.details}\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n\u003Csummary>Advanced Techniques\u003C\u002Fsummary>\n${progressivePrompt.advanced}\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n\u003Csummary>Examples\u003C\u002Fsummary>\n${progressivePrompt.examples}\n\u003C\u002Fdetails>\n`;\n```\n\n### Validation Criteria\n- [ ] All 7 components present\n- [ ] Logical flow established\n- [ ] Progressive disclosure applied\n- [ ] Clear hierarchy visible\n\n### Script Template\n```bash\n#!\u002Fbin\u002Fbash\n# optimize-structure.sh\n\nINPUT_PROMPT=\"$1\"\nOUTPUT_PROMPT=\"$2\"\n\n# Analyze structure\nANALYSIS=$(npx claude-flow@alpha agent-spawn \\\n  --type researcher \\\n  --task \"Analyze prompt structure: $(cat $INPUT_PROMPT)\")\n\n# Optimize\nOPTIMIZED=$(npx claude-flow@alpha agent-spawn \\\n  --type coder \\\n  --task \"Restructure prompt based on analysis: $ANALYSIS\")\n\necho \"$OPTIMIZED\" > \"$OUTPUT_PROMPT\"\n\nnpx claude-flow@alpha hooks post-edit \\\n  --file \"$OUTPUT_PROMPT\" \\\n  --memory-key \"prompt-architect\u002Foptimized\"\n```\n\n## Phase 3: Apply Evidence-Based Techniques\n\n### Objective\nIncorporate proven prompt engineering methods\n\n### Agent: Researcher + Coder\n\n**Step 3.1: Add Chain-of-Thought**\n```markdown\n# Chain-of-Thought Enhancement\n\nBefore providing your final answer, think through the problem step by step:\n\n1. **Understand**: Restate the problem in your own words\n2. **Analyze**: Break down into components\n3. **Reason**: Work through the logic\n4. **Synthesize**: Combine insights\n5. **Conclude**: Provide final answer\n\nFormat your response as:\n\u003Cthinking>\n[Your step-by-step reasoning]\n\u003C\u002Fthinking>\n\n\u003Canswer>\n[Your final answer]\n\u003C\u002Fanswer>\n```\n\n**Step 3.2: Add Self-Consistency**\n```markdown\n# Self-Consistency Pattern\n\nGenerate 3 independent solutions to this problem using different approaches:\n\nApproach 1: [Method 1]\nApproach 2: [Method 2]\nApproach 3: [Method 3]\n\nThen compare the solutions and select the most robust answer, explaining why it's superior.\n```\n\n**Step 3.3: Add ReAct Pattern**\n```markdown\n# ReAct (Reasoning + Acting) Pattern\n\nFor each step in your process:\n\n**Thought**: [What you're thinking]\n**Action**: [What you're doing]\n**Observation**: [What you learned]\n\nRepeat this cycle until the task is complete.\n\nExample:\nThought: I need to understand the data structure\nAction: Analyze the schema\nObservation: It's a relational database with 5 tables\nThought: I should check for relationships\nAction: Examine foreign keys\nObservation: Tables are connected via user_id\n```\n\n**Step 3.4: Add Few-Shot Examples**\n```javascript\nconst examples = [\n  {\n    input: '[Example input 1]',\n    reasoning: '[How to approach it]',\n    output: '[Expected output 1]'\n  },\n  {\n    input: '[Example input 2]',\n    reasoning: '[How to approach it]',\n    output: '[Expected output 2]'\n  },\n  {\n    input: '[Example input 3 - edge case]',\n    reasoning: '[How to handle edge case]',\n    output: '[Expected output 3]'\n  }\n];\n\nconst fewShotSection = `\n# Examples\n\n${examples.map((ex, i) => `\n## Example ${i + 1}\n\n**Input**: ${ex.input}\n\n**Reasoning**: ${ex.reasoning}\n\n**Output**:\n\\`\\`\\`\n${ex.output}\n\\`\\`\\`\n`).join('\\n')}\n\nNow apply the same pattern to: [ACTUAL INPUT]\n`;\n```\n\n**Step 3.5: Add Constraint Framing**\n```markdown\n# Constraint-Based Optimization\n\nThis task requires balancing multiple constraints:\n\n1. **Quality**: Must meet 90% accuracy\n2. **Speed**: Must complete in \u003C 2 seconds\n3. **Resources**: Memory usage \u003C 100MB\n4. **Safety**: No external API calls\n\nWhen these constraints conflict, prioritize in this order: Safety > Quality > Speed > Resources\n\nIf you cannot satisfy all constraints, explicitly state which ones are violated and why.\n```\n\n### Validation Criteria\n- [ ] CoT reasoning added\n- [ ] Self-consistency pattern included\n- [ ] Examples provided (2-3)\n- [ ] Constraints clearly framed\n\n### Memory Pattern\n```bash\nnpx claude-flow@alpha hooks post-edit \\\n  --file \"enhanced-prompt.md\" \\\n  --memory-key \"prompt-architect\u002Ftechniques-applied\"\n```\n\n## Phase 4: Validate Effectiveness\n\n### Objective\nTest prompt performance and measure improvement\n\n### Agent: Researcher\n\n**Step 4.1: Define Test Cases**\n```javascript\nconst testCases = [\n  {\n    id: 'test-1',\n    type: 'typical',\n    input: '[Common use case]',\n    expectedOutput: '[What should happen]',\n    successCriteria: '[How to measure success]'\n  },\n  {\n    id: 'test-2',\n    type: 'edge-case',\n    input: '[Unusual scenario]',\n    expectedOutput: '[How to handle]',\n    successCriteria: '[Validation method]'\n  },\n  {\n    id: 'test-3',\n    type: 'stress',\n    input: '[Complex\u002Fambiguous input]',\n    expectedOutput: '[Robust handling]',\n    successCriteria: '[Quality threshold]'\n  }\n];\n\nawait memory.store('prompt-architect\u002Ftest-cases', testCases);\n```\n\n**Step 4.2: Run A\u002FB Tests**\n```javascript\nasync function runABTest(originalPrompt, optimizedPrompt, testCases) {\n  const results = {\n    original: [],\n    optimized: []\n  };\n\n  for (const testCase of testCases) {\n    \u002F\u002F Test original prompt\n    const originalResult = await testPrompt(originalPrompt, testCase.input);\n    results.original.push({\n      testId: testCase.id,\n      output: originalResult,\n      score: scoreOutput(originalResult, testCase.expectedOutput),\n      meetsSuccessCriteria: evaluateCriteria(originalResult, testCase.successCriteria)\n    });\n\n    \u002F\u002F Test optimized prompt\n    const optimizedResult = await testPrompt(optimizedPrompt, testCase.input);\n    results.optimized.push({\n      testId: testCase.id,\n      output: optimizedResult,\n      score: scoreOutput(optimizedResult, testCase.expectedOutput),\n      meetsSuccessCriteria: evaluateCriteria(optimizedResult, testCase.successCriteria)\n    });\n  }\n\n  return results;\n}\n\nconst abTestResults = await runABTest(originalPrompt, optimizedPrompt, testCases);\nawait memory.store('prompt-architect\u002Fab-test-results', abTestResults);\n```\n\n**Step 4.3: Calculate Metrics**\n```javascript\nconst metrics = {\n  original: {\n    avgScore: calculateAverage(abTestResults.original.map(r => r.score)),\n    successRate: calculateSuccessRate(abTestResults.original),\n    consistency: calculateConsistency(abTestResults.original)\n  },\n  optimized: {\n    avgScore: calculateAverage(abTestResults.optimized.map(r => r.score)),\n    successRate: calculateSuccessRate(abTestResults.optimized),\n    consistency: calculateConsistency(abTestResults.optimized)\n  },\n  improvement: {\n    scoreImprovement: 0, \u002F\u002F calculated below\n    successRateImprovement: 0,\n    consistencyImprovement: 0\n  }\n};\n\nmetrics.improvement = {\n  scoreImprovement: ((metrics.optimized.avgScore - metrics.original.avgScore) \u002F metrics.original.avgScore * 100).toFixed(2) + '%',\n  successRateImprovement: ((metrics.optimized.successRate - metrics.original.successRate) \u002F metrics.original.successRate * 100).toFixed(2) + '%',\n  consistencyImprovement: ((metrics.optimized.consistency - metrics.original.consistency) \u002F metrics.original.consistency * 100).toFixed(2) + '%'\n};\n\nawait memory.store('prompt-architect\u002Fmetrics', metrics);\n```\n\n### Validation Criteria\n- [ ] Test cases defined (3+ cases)\n- [ ] A\u002FB testing completed\n- [ ] Metrics calculated\n- [ ] Improvement demonstrated\n\n### Script Template\n```bash\n#!\u002Fbin\u002Fbash\n# validate-prompt.sh\n\nORIGINAL=\"$1\"\nOPTIMIZED=\"$2\"\nTEST_CASES=\"$3\"\n\n# Run A\u002FB tests\nRESULTS=$(npx claude-flow@alpha agent-spawn \\\n  --type researcher \\\n  --task \"A\u002FB test prompts: original='$ORIGINAL' optimized='$OPTIMIZED' tests='$TEST_CASES'\")\n\n# Calculate improvement\nMETRICS=$(echo \"$RESULTS\" | jq '.improvement')\n\necho \"Improvement Metrics:\"\necho \"$METRICS\" | jq '.'\n\n# Store results\nnpx claude-flow@alpha hooks post-task \\\n  --task-id \"prompt-validation\" \\\n  --metrics \"$METRICS\"\n```\n\n## Phase 5: Refine Iteratively\n\n### Objective\nContinuous improvement based on validation results\n\n### Agent: Coder\n\n**Step 5.1: Analyze Failures**\n```javascript\nconst failures = abTestResults.optimized.filter(r => !r.meetsSuccessCriteria);\n\nconst failureAnalysis = failures.map(failure => {\n  const testCase = testCases.find(tc => tc.id === failure.testId);\n\n  return {\n    testId: failure.testId,\n    testType: testCase.type,\n    expectedOutput: testCase.expectedOutput,\n    actualOutput: failure.output,\n    gap: analyzeGap(testCase.expectedOutput, failure.output),\n    rootCause: identifyRootCause(failure, optimizedPrompt),\n    recommendedFix: suggestFix(rootCause)\n  };\n});\n\nawait memory.store('prompt-architect\u002Ffailure-analysis', failureAnalysis);\n```\n\n**Step 5.2: Apply Refinements**\n```javascript\nlet refinedPrompt = optimizedPrompt;\n\nfor (const analysis of failureAnalysis) {\n  switch (analysis.rootCause.type) {\n    case 'MISSING_CONSTRAINT':\n      refinedPrompt = addConstraint(refinedPrompt, analysis.recommendedFix.constraint);\n      break;\n    case 'AMBIGUOUS_INSTRUCTION':\n      refinedPrompt = clarifyInstruction(refinedPrompt, analysis.recommendedFix.clarification);\n      break;\n    case 'INSUFFICIENT_EXAMPLES':\n      refinedPrompt = addExample(refinedPrompt, analysis.recommendedFix.example);\n      break;\n    case 'MISSING_EDGE_CASE_HANDLING':\n      refinedPrompt = addEdgeCaseGuidance(refinedPrompt, analysis.recommendedFix.guidance);\n      break;\n  }\n}\n\nawait memory.store('prompt-architect\u002Frefined-prompt', refinedPrompt);\n```\n\n**Step 5.3: Re-validate**\n```javascript\nconst revalidationResults = await runABTest(optimizedPrompt, refinedPrompt, testCases);\n\nconst improvementFromOptimized = {\n  scoreImprovement: calculateImprovement(\n    metrics.optimized.avgScore,\n    calculateAverage(revalidationResults.optimized.map(r => r.score))\n  ),\n  successRateImprovement: calculateImprovement(\n    metrics.optimized.successRate,\n    calculateSuccessRate(revalidationResults.optimized)\n  )\n};\n\n\u002F\u002F If refined version is better, adopt it\nif (improvementFromOptimized.scoreImprovement > 5) {\n  console.log('✅ Refinement successful. Adopting refined version.');\n  finalPrompt = refinedPrompt;\n} else {\n  console.log('✅ Optimized version is sufficient.');\n  finalPrompt = optimizedPrompt;\n}\n\nawait memory.store('prompt-architect\u002Ffinal-prompt', finalPrompt);\n```\n\n**Step 5.4: Generate Documentation**\n```markdown\n# Prompt Optimization Report\n\n## Original Prompt\n\\`\\`\\`\n${originalPrompt}\n\\`\\`\\`\n\n## Final Prompt\n\\`\\`\\`\n${finalPrompt}\n\\`\\`\\`\n\n## Changes Applied\n${changesApplied.map(change => `- ${change}`).join('\\n')}\n\n## Performance Metrics\n\n| Metric | Original | Optimized | Improvement |\n|--------|----------|-----------|-------------|\n| Avg Score | ${metrics.original.avgScore} | ${metrics.optimized.avgScore} | ${metrics.improvement.scoreImprovement} |\n| Success Rate | ${metrics.original.successRate}% | ${metrics.optimized.successRate}% | ${metrics.improvement.successRateImprovement} |\n| Consistency | ${metrics.original.consistency} | ${metrics.optimized.consistency} | ${metrics.improvement.consistencyImprovement} |\n\n## Test Results\n${testCases.map(tc => `\n### ${tc.id} (${tc.type})\n- **Expected**: ${tc.expectedOutput}\n- **Original Result**: ${getResult('original', tc.id)}\n- **Optimized Result**: ${getResult('optimized', tc.id)}\n- **Status**: ${getStatus('optimized', tc.id)}\n`).join('\\n')}\n\n## Recommendations\n${generateRecommendations(finalPrompt)}\n```\n\n### Validation Criteria\n- [ ] Failures analyzed\n- [ ] Refinements applied\n- [ ] Re-validation completed\n- [ ] Documentation generated\n\n### Memory Pattern\n```bash\nnpx claude-flow@alpha hooks session-end \\\n  --session-id \"prompt-architect-${TIMESTAMP}\" \\\n  --export-metrics true \\\n  --summary \"Prompt optimized with ${IMPROVEMENT}% improvement\"\n```\n\n## Success Metrics\n\n### Quantitative\n- Score improvement > 20%\n- Success rate > 85%\n- Consistency score > 0.8\n- All test cases pass\n\n### Qualitative\n- Clear structure\n- No anti-patterns\n- Evidence-based techniques applied\n- User satisfaction with outputs\n\n## Common Patterns\n\n### Pattern 1: Role-Based Optimization\n```markdown\nYou are an expert ${domain} specialist with ${years} years of experience.\nYour expertise includes: ${expertise_list}\nYou approach problems by: ${methodology}\n```\n\n### Pattern 2: Constraint-First Design\n```markdown\n# Constraints (Read First)\nMUST: ${required_constraints}\nMUST NOT: ${forbidden_actions}\nOPTIMIZE FOR: ${optimization_targets}\n\n# Task\n[task description with constraints in mind]\n```\n\n### Pattern 3: Format-Driven Output\n```markdown\nYour output MUST follow this exact structure:\n\n\\`\\`\\`json\n{\n  \"analysis\": \"...\",\n  \"recommendations\": [...],\n  \"confidence\": 0.XX\n}\n\\`\\`\\`\n\nDo not deviate from this format.\n```\n\n## Integration Examples\n\n### With Agent Creation\n```javascript\n\u002F\u002F Use prompt architect to optimize agent system prompts\nconst agentPrompt = await optimizePrompt({\n  role: 'backend-developer',\n  domain: 'Node.js API development',\n  constraints: ['RESTful design', 'security-first', 'test-driven'],\n  outputFormat: 'production-ready code'\n});\n```\n\n### With SPARC Workflow\n```bash\n# Optimize specification prompts\nnpx claude-flow@alpha skill-run prompt-architect \\\n  --input \"sparc-spec-prompt.md\" \\\n  --output \"optimized-spec-prompt.md\"\n\n# Use optimized prompt in SPARC\nnpx claude-flow@alpha sparc run spec-pseudocode \\\n  --prompt-file \"optimized-spec-prompt.md\"\n```\n\n## Memory Schema\n\n```javascript\n{\n  \"prompt-architect\u002F\": {\n    \"session-${id}\u002F\": {\n      \"analysis\": {},\n      \"anti-patterns\": [],\n      \"missing\": [],\n      \"structure\": {},\n      \"techniques-applied\": [],\n      \"test-cases\": [],\n      \"ab-test-results\": {},\n      \"metrics\": {},\n      \"failure-analysis\": [],\n      \"refined-prompt\": \"\",\n      \"final-prompt\": \"\"\n    }\n  }\n}\n```\n\n## Skill Completion\n\nOutputs:\n1. **final-prompt.md**: Optimized prompt ready for use\n2. **optimization-report.md**: Detailed analysis and results\n3. **ab-test-results.json**: Performance comparison data\n4. **prompt-library-entry.md**: Cataloged for reuse\n\nSkill complete when metrics show >20% improvement and all test cases pass.\n","# Prompt Architect - Evidence-Based Prompt Engineering\n\n## Overview\n\nComprehensive framework for analyzing, creating, and refining prompts for AI systems (Claude, GPT, etc.). Applies structural optimization, self-consistency patterns, and anti-pattern detection to transform prompts into highly effective versions.\n\n## When to Use This Skill\n\n- Creating new prompts for AI systems\n- Existing prompts produce poor results\n- Inconsistent AI outputs\n- Need to improve prompt clarity\n- Applying evidence-based prompt engineering\n- Optimizing agent instructions\n- Building prompt libraries\n\n## Theoretical Foundation\n\n### Evidence-Based Techniques\n\n1. **Chain-of-Thought (CoT)**: Explicit reasoning steps\n2. **Self-Consistency**: Multiple reasoning paths\n3. **ReAct**: Reasoning + Acting pattern\n4. **Program-of-Thought**: Structured logic\n5. **Plan-and-Solve**: Decomposition strategy\n6. **Role-Playing**: Persona assignment\n7. **Few-Shot Learning**: Example-based instruction\n\n### Prompt Structure Principles\n\n```\n[System Context] → [Role Definition] → [Task Description] →\n[Constraints] → [Format Specification] → [Examples] → [Quality Criteria]\n```\n\n## Phase 1: Analyze Current Prompt\n\n### Objective\nIdentify weaknesses and improvement opportunities\n\n### Agent: Researcher\n\n**Step 1.1: Structural Analysis**\n```javascript\nconst promptAnalysis = {\n  components: {\n    hasSystemContext: checkForContext(prompt),\n    hasRoleDefinition: checkForRole(prompt),\n    hasTaskDescription: checkForTask(prompt),\n    hasConstraints: checkForConstraints(prompt),\n    hasFormatSpec: checkForFormat(prompt),\n    hasExamples: checkForExamples(prompt),\n    hasQualityCriteria: checkForCriteria(prompt)\n  },\n  metrics: {\n    length: prompt.length,\n    clarity: calculateClarity(prompt),\n    specificity: calculateSpecificity(prompt),\n    completeness: calculateCompleteness(prompt)\n  },\n  antiPatterns: detectAntiPatterns(prompt)\n};\n\nawait memory.store('prompt-architect\u002Fanalysis', promptAnalysis);\n```\n\n**Step 1.2: Detect Anti-Patterns**\n```javascript\nconst antiPatterns = [\n  {\n    name: 'Vague Instructions',\n    pattern: \u002Fplease|try to|maybe|possibly\u002Fgi,\n    severity: 'HIGH',\n    fix: 'Use imperative commands: \"Analyze...\", \"Generate...\", \"Create...\"'\n  },\n  {\n    name: 'Missing Context',\n    pattern: absence of background info,\n    severity: 'HIGH',\n    fix: 'Add system context and domain information'\n  },\n  {\n    name: 'No Output Format',\n    pattern: absence of format specification,\n    severity: 'MEDIUM',\n    fix: 'Specify exact output format (JSON, markdown, etc.)'\n  },\n  {\n    name: 'Conflicting Instructions',\n    pattern: detectContradictions(prompt),\n    severity: 'HIGH',\n    fix: 'Resolve contradictions, prioritize requirements'\n  },\n  {\n    name: 'Implicit Assumptions',\n    pattern: detectImplicitAssumptions(prompt),\n    severity: 'MEDIUM',\n    fix: 'Make all assumptions explicit'\n  }\n];\n\nconst foundAntiPatterns = antiPatterns.filter(ap =>\n  ap.pattern.test ? ap.pattern.test(prompt) : ap.pattern\n);\n\nawait memory.store('prompt-architect\u002Fanti-patterns', foundAntiPatterns);\n```\n\n**Step 1.3: Identify Missing Components**\n```javascript\nconst missingComponents = [];\n\nif (!promptAnalysis.components.hasSystemContext) {\n  missingComponents.push({\n    component: 'System Context',\n    importance: 'HIGH',\n    recommendation: 'Add background info, domain knowledge, constraints'\n  });\n}\n\nif (!promptAnalysis.components.hasExamples) {\n  missingComponents.push({\n    component: 'Examples',\n    importance: 'MEDIUM',\n    recommendation: 'Add 2-3 examples showing desired behavior'\n  });\n}\n\n\u002F\u002F ... check other components\n\nawait memory.store('prompt-architect\u002Fmissing', missingComponents);\n```\n\n### Validation Criteria\n- [ ] All 7 components checked\n- [ ] Anti-patterns identified\n- [ ] Missing components listed\n- [ ] Severity assigned to issues\n\n### Hooks Integration\n```bash\nnpx claude-flow@alpha hooks pre-task \\\n  --description \"Analyze prompt structure and quality\" \\\n  --complexity \"medium\"\n\nnpx claude-flow@alpha hooks post-task \\\n  --task-id \"prompt-analysis\" \\\n  --output \"analysis-report.json\"\n```\n\n## Phase 2: Structure Optimization\n\n### Objective\nReorganize prompt for logical flow and clarity\n\n### Agent: Coder (Prompt Specialist)\n\n**Step 2.1: Apply Template Structure**\n```javascript\nconst optimizedStructure = {\n  systemContext: {\n    domain: extractDomain(prompt),\n    background: generateContextualBackground(),\n    constraints: extractOrInferConstraints(prompt)\n  },\n  roleDefinition: {\n    persona: definePersona(task),\n    expertise: listRequiredExpertise(),\n    perspective: defineWorkingPerspective()\n  },\n  taskDescription: {\n    primary: extractPrimaryTask(prompt),\n    secondary: extractSecondaryTasks(prompt),\n    scope: defineScope(),\n    outOfScope: defineWhatNotToDo()\n  },\n  constraints: {\n    required: extractRequirements(prompt),\n    forbidden: extractProhibitions(prompt),\n    optional: extractPreferences(prompt)\n  },\n  formatSpecification: {\n    outputFormat: specifyFormat(),\n    structure: defineStructure(),\n    examples: []\n  },\n  qualityCriteria: {\n    success: defineSuccessCriteria(),\n    validation: defineValidationMethod(),\n    metrics: defineMetrics()\n  }\n};\n\nawait memory.store('prompt-architect\u002Fstructure', optimizedStructure);\n```\n\n**Step 2.2: Build Optimized Prompt**\n```markdown\n# System Context\n${optimizedStructure.systemContext.background}\n\nDomain: ${optimizedStructure.systemContext.domain}\nConstraints: ${optimizedStructure.systemContext.constraints.join(', ')}\n\n# Role\nYou are ${optimizedStructure.roleDefinition.persona} with expertise in:\n${optimizedStructure.roleDefinition.expertise.map(e => `- ${e}`).join('\\n')}\n\nYour perspective: ${optimizedStructure.roleDefinition.perspective}\n\n# Task\nPrimary objective: ${optimizedStructure.taskDescription.primary}\n\n${optimizedStructure.taskDescription.secondary.length > 0 ?\n  'Secondary objectives:\\n' + optimizedStructure.taskDescription.secondary.map(t => `- ${t}`).join('\\n') : ''}\n\nScope: ${optimizedStructure.taskDescription.scope}\nOut of scope: ${optimizedStructure.taskDescription.outOfScope.join(', ')}\n\n# Constraints\nMUST: ${optimizedStructure.constraints.required.map(r => `- ${r}`).join('\\n')}\nMUST NOT: ${optimizedStructure.constraints.forbidden.map(f => `- ${f}`).join('\\n')}\nPREFER: ${optimizedStructure.constraints.optional.map(o => `- ${o}`).join('\\n')}\n\n# Output Format\n${optimizedStructure.formatSpecification.outputFormat}\n\nStructure:\n${optimizedStructure.formatSpecification.structure}\n\n# Quality Criteria\nSuccess is defined as:\n${optimizedStructure.qualityCriteria.success.map(s => `- ${s}`).join('\\n')}\n\nValidation method: ${optimizedStructure.qualityCriteria.validation}\n```\n\n**Step 2.3: Add Progressive Disclosure**\n```javascript\n\u002F\u002F For complex prompts, use hierarchical structure\nconst progressivePrompt = {\n  essential: generateEssentialInstructions(),\n  details: generateDetailedGuidance(),\n  advanced: generateAdvancedTechniques(),\n  examples: generateExamples(),\n  troubleshooting: generateTroubleshootingGuide()\n};\n\n\u002F\u002F Structure with collapsible sections\nconst enhancedPrompt = `\n${progressivePrompt.essential}\n\n\u003Cdetails>\n\u003Csummary>Detailed Guidance\u003C\u002Fsummary>\n${progressivePrompt.details}\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n\u003Csummary>Advanced Techniques\u003C\u002Fsummary>\n${progressivePrompt.advanced}\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n\u003Csummary>Examples\u003C\u002Fsummary>\n${progressivePrompt.examples}\n\u003C\u002Fdetails>\n`;\n```\n\n### Validation Criteria\n- [ ] All 7 components present\n- [ ] Logical flow established\n- [ ] Progressive disclosure applied\n- [ ] Clear hierarchy visible\n\n### Script Template\n```bash\n#!\u002Fbin\u002Fbash\n# optimize-structure.sh\n\nINPUT_PROMPT=\"$1\"\nOUTPUT_PROMPT=\"$2\"\n\n# Analyze structure\nANALYSIS=$(npx claude-flow@alpha agent-spawn \\\n  --type researcher \\\n  --task \"Analyze prompt structure: $(cat $INPUT_PROMPT)\")\n\n# Optimize\nOPTIMIZED=$(npx claude-flow@alpha agent-spawn \\\n  --type coder \\\n  --task \"Restructure prompt based on analysis: $ANALYSIS\")\n\necho \"$OPTIMIZED\" > \"$OUTPUT_PROMPT\"\n\nnpx claude-flow@alpha hooks post-edit \\\n  --file \"$OUTPUT_PROMPT\" \\\n  --memory-key \"prompt-architect\u002Foptimized\"\n```\n\n## Phase 3: Apply Evidence-Based Techniques\n\n### Objective\nIncorporate proven prompt engineering methods\n\n### Agent: Researcher + Coder\n\n**Step 3.1: Add Chain-of-Thought**\n```markdown\n# Chain-of-Thought Enhancement\n\nBefore providing your final answer, think through the problem step by step:\n\n1. **Understand**: Restate the problem in your own words\n2. **Analyze**: Break down into components\n3. **Reason**: Work through the logic\n4. **Synthesize**: Combine insights\n5. **Conclude**: Provide final answer\n\nFormat your response as:\n\u003Cthinking>\n[Your step-by-step reasoning]\n\u003C\u002Fthinking>\n\n\u003Canswer>\n[Your final answer]\n\u003C\u002Fanswer>\n```\n\n**Step 3.2: Add Self-Consistency**\n```markdown\n# Self-Consistency Pattern\n\nGenerate 3 independent solutions to this problem using different approaches:\n\nApproach 1: [Method 1]\nApproach 2: [Method 2]\nApproach 3: [Method 3]\n\nThen compare the solutions and select the most robust answer, explaining why it's superior.\n```\n\n**Step 3.3: Add ReAct Pattern**\n```markdown\n# ReAct (Reasoning + Acting) Pattern\n\nFor each step in your process:\n\n**Thought**: [What you're thinking]\n**Action**: [What you're doing]\n**Observation**: [What you learned]\n\nRepeat this cycle until the task is complete.\n\nExample:\nThought: I need to understand the data structure\nAction: Analyze the schema\nObservation: It's a relational database with 5 tables\nThought: I should check for relationships\nAction: Examine foreign keys\nObservation: Tables are connected via user_id\n```\n\n**Step 3.4: Add Few-Shot Examples**\n```javascript\nconst examples = [\n  {\n    input: '[Example input 1]',\n    reasoning: '[How to approach it]',\n    output: '[Expected output 1]'\n  },\n  {\n    input: '[Example input 2]',\n    reasoning: '[How to approach it]',\n    output: '[Expected output 2]'\n  },\n  {\n    input: '[Example input 3 - edge case]',\n    reasoning: '[How to handle edge case]',\n    output: '[Expected output 3]'\n  }\n];\n\nconst fewShotSection = `\n# Examples\n\n${examples.map((ex, i) => `\n## Example ${i + 1}\n\n**Input**: ${ex.input}\n\n**Reasoning**: ${ex.reasoning}\n\n**Output**:\n\\`\\`\\`\n${ex.output}\n\\`\\`\\`\n`).join('\\n')}\n\nNow apply the same pattern to: [ACTUAL INPUT]\n`;\n```\n\n**Step 3.5: Add Constraint Framing**\n```markdown\n# Constraint-Based Optimization\n\nThis task requires balancing multiple constraints:\n\n1. **Quality**: Must meet 90% accuracy\n2. **Speed**: Must complete in \u003C 2 seconds\n3. **Resources**: Memory usage \u003C 100MB\n4. **Safety**: No external API calls\n\nWhen these constraints conflict, prioritize in this order: Safety > Quality > Speed > Resources\n\nIf you cannot satisfy all constraints, explicitly state which ones are violated and why.\n```\n\n### Validation Criteria\n- [ ] CoT reasoning added\n- [ ] Self-consistency pattern included\n- [ ] Examples provided (2-3)\n- [ ] Constraints clearly framed\n\n### Memory Pattern\n```bash\nnpx claude-flow@alpha hooks post-edit \\\n  --file \"enhanced-prompt.md\" \\\n  --memory-key \"prompt-architect\u002Ftechniques-applied\"\n```\n\n## Phase 4: Validate Effectiveness\n\n### Objective\nTest prompt performance and measure improvement\n\n### Agent: Researcher\n\n**Step 4.1: Define Test Cases**\n```javascript\nconst testCases = [\n  {\n    id: 'test-1',\n    type: 'typical',\n    input: '[Common use case]',\n    expectedOutput: '[What should happen]',\n    successCriteria: '[How to measure success]'\n  },\n  {\n    id: 'test-2',\n    type: 'edge-case',\n    input: '[Unusual scenario]',\n    expectedOutput: '[How to handle]',\n    successCriteria: '[Validation method]'\n  },\n  {\n    id: 'test-3',\n    type: 'stress',\n    input: '[Complex\u002Fambiguous input]',\n    expectedOutput: '[Robust handling]',\n    successCriteria: '[Quality threshold]'\n  }\n];\n\nawait memory.store('prompt-architect\u002Ftest-cases', testCases);\n```\n\n**Step 4.2: Run A\u002FB Tests**\n```javascript\nasync function runABTest(originalPrompt, optimizedPrompt, testCases) {\n  const results = {\n    original: [],\n    optimized: []\n  };\n\n  for (const testCase of testCases) {\n    \u002F\u002F Test original prompt\n    const originalResult = await testPrompt(originalPrompt, testCase.input);\n    results.original.push({\n      testId: testCase.id,\n      output: originalResult,\n      score: scoreOutput(originalResult, testCase.expectedOutput),\n      meetsSuccessCriteria: evaluateCriteria(originalResult, testCase.successCriteria)\n    });\n\n    \u002F\u002F Test optimized prompt\n    const optimizedResult = await testPrompt(optimizedPrompt, testCase.input);\n    results.optimized.push({\n      testId: testCase.id,\n      output: optimizedResult,\n      score: scoreOutput(optimizedResult, testCase.expectedOutput),\n      meetsSuccessCriteria: evaluateCriteria(optimizedResult, testCase.successCriteria)\n    });\n  }\n\n  return results;\n}\n\nconst abTestResults = await runABTest(originalPrompt, optimizedPrompt, testCases);\nawait memory.store('prompt-architect\u002Fab-test-results', abTestResults);\n```\n\n**Step 4.3: Calculate Metrics**\n```javascript\nconst metrics = {\n  original: {\n    avgScore: calculateAverage(abTestResults.original.map(r => r.score)),\n    successRate: calculateSuccessRate(abTestResults.original),\n    consistency: calculateConsistency(abTestResults.original)\n  },\n  optimized: {\n    avgScore: calculateAverage(abTestResults.optimized.map(r => r.score)),\n    successRate: calculateSuccessRate(abTestResults.optimized),\n    consistency: calculateConsistency(abTestResults.optimized)\n  },\n  improvement: {\n    scoreImprovement: 0, \u002F\u002F calculated below\n    successRateImprovement: 0,\n    consistencyImprovement: 0\n  }\n};\n\nmetrics.improvement = {\n  scoreImprovement: ((metrics.optimized.avgScore - metrics.original.avgScore) \u002F metrics.original.avgScore * 100).toFixed(2) + '%',\n  successRateImprovement: ((metrics.optimized.successRate - metrics.original.successRate) \u002F metrics.original.successRate * 100).toFixed(2) + '%',\n  consistencyImprovement: ((metrics.optimized.consistency - metrics.original.consistency) \u002F metrics.original.consistency * 100).toFixed(2) + '%'\n};\n\nawait memory.store('prompt-architect\u002Fmetrics', metrics);\n```\n\n### Validation Criteria\n- [ ] Test cases defined (3+ cases)\n- [ ] A\u002FB testing completed\n- [ ] Metrics calculated\n- [ ] Improvement demonstrated\n\n### Script Template\n```bash\n#!\u002Fbin\u002Fbash\n# validate-prompt.sh\n\nORIGINAL=\"$1\"\nOPTIMIZED=\"$2\"\nTEST_CASES=\"$3\"\n\n# Run A\u002FB tests\nRESULTS=$(npx claude-flow@alpha agent-spawn \\\n  --type researcher \\\n  --task \"A\u002FB test prompts: original='$ORIGINAL' optimized='$OPTIMIZED' tests='$TEST_CASES'\")\n\n# Calculate improvement\nMETRICS=$(echo \"$RESULTS\" | jq '.improvement')\n\necho \"Improvement Metrics:\"\necho \"$METRICS\" | jq '.'\n\n# Store results\nnpx claude-flow@alpha hooks post-task \\\n  --task-id \"prompt-validation\" \\\n  --metrics \"$METRICS\"\n```\n\n## Phase 5: Refine Iteratively\n\n### Objective\nContinuous improvement based on validation results\n\n### Agent: Coder\n\n**Step 5.1: Analyze Failures**\n```javascript\nconst failures = abTestResults.optimized.filter(r => !r.meetsSuccessCriteria);\n\nconst failureAnalysis = failures.map(failure => {\n  const testCase = testCases.find(tc => tc.id === failure.testId);\n\n  return {\n    testId: failure.testId,\n    testType: testCase.type,\n    expectedOutput: testCase.expectedOutput,\n    actualOutput: failure.output,\n    gap: analyzeGap(testCase.expectedOutput, failure.output),\n    rootCause: identifyRootCause(failure, optimizedPrompt),\n    recommendedFix: suggestFix(rootCause)\n  };\n});\n\nawait memory.store('prompt-architect\u002Ffailure-analysis', failureAnalysis);\n```\n\n**Step 5.2: Apply Refinements**\n```javascript\nlet refinedPrompt = optimizedPrompt;\n\nfor (const analysis of failureAnalysis) {\n  switch (analysis.rootCause.type) {\n    case 'MISSING_CONSTRAINT':\n      refinedPrompt = addConstraint(refinedPrompt, analysis.recommendedFix.constraint);\n      break;\n    case 'AMBIGUOUS_INSTRUCTION':\n      refinedPrompt = clarifyInstruction(refinedPrompt, analysis.recommendedFix.clarification);\n      break;\n    case 'INSUFFICIENT_EXAMPLES':\n      refinedPrompt = addExample(refinedPrompt, analysis.recommendedFix.example);\n      break;\n    case 'MISSING_EDGE_CASE_HANDLING':\n      refinedPrompt = addEdgeCaseGuidance(refinedPrompt, analysis.recommendedFix.guidance);\n      break;\n  }\n}\n\nawait memory.store('prompt-architect\u002Frefined-prompt', refinedPrompt);\n```\n\n**Step 5.3: Re-validate**\n```javascript\nconst revalidationResults = await runABTest(optimizedPrompt, refinedPrompt, testCases);\n\nconst improvementFromOptimized = {\n  scoreImprovement: calculateImprovement(\n    metrics.optimized.avgScore,\n    calculateAverage(revalidationResults.optimized.map(r => r.score))\n  ),\n  successRateImprovement: calculateImprovement(\n    metrics.optimized.successRate,\n    calculateSuccessRate(revalidationResults.optimized)\n  )\n};\n\n\u002F\u002F If refined version is better, adopt it\nif (improvementFromOptimized.scoreImprovement > 5) {\n  console.log('✅ Refinement successful. Adopting refined version.');\n  finalPrompt = refinedPrompt;\n} else {\n  console.log('✅ Optimized version is sufficient.');\n  finalPrompt = optimizedPrompt;\n}\n\nawait memory.store('prompt-architect\u002Ffinal-prompt', finalPrompt);\n```\n\n**Step 5.4: Generate Documentation**\n```markdown\n# Prompt Optimization Report\n\n## Original Prompt\n\\`\\`\\`\n${originalPrompt}\n\\`\\`\\`\n\n## Final Prompt\n\\`\\`\\`\n${finalPrompt}\n\\`\\`\\`\n\n## Changes Applied\n${changesApplied.map(change => `- ${change}`).join('\\n')}\n\n## Performance Metrics\n\n| Metric | Original | Optimized | Improvement |\n|--------|----------|-----------|-------------|\n| Avg Score | ${metrics.original.avgScore} | ${metrics.optimized.avgScore} | ${metrics.improvement.scoreImprovement} |\n| Success Rate | ${metrics.original.successRate}% | ${metrics.optimized.successRate}% | ${metrics.improvement.successRateImprovement} |\n| Consistency | ${metrics.original.consistency} | ${metrics.optimized.consistency} | ${metrics.improvement.consistencyImprovement} |\n\n## Test Results\n${testCases.map(tc => `\n### ${tc.id} (${tc.type})\n- **Expected**: ${tc.expectedOutput}\n- **Original Result**: ${getResult('original', tc.id)}\n- **Optimized Result**: ${getResult('optimized', tc.id)}\n- **Status**: ${getStatus('optimized', tc.id)}\n`).join('\\n')}\n\n## Recommendations\n${generateRecommendations(finalPrompt)}\n```\n\n### Validation Criteria\n- [ ] Failures analyzed\n- [ ] Refinements applied\n- [ ] Re-validation completed\n- [ ] Documentation generated\n\n### Memory Pattern\n```bash\nnpx claude-flow@alpha hooks session-end \\\n  --session-id \"prompt-architect-${TIMESTAMP}\" \\\n  --export-metrics true \\\n  --summary \"Prompt optimized with ${IMPROVEMENT}% improvement\"\n```\n\n## Success Metrics\n\n### Quantitative\n- Score improvement > 20%\n- Success rate > 85%\n- Consistency score > 0.8\n- All test cases pass\n\n### Qualitative\n- Clear structure\n- No anti-patterns\n- Evidence-based techniques applied\n- User satisfaction with outputs\n\n## Common Patterns\n\n### Pattern 1: Role-Based Optimization\n```markdown\nYou are an expert ${domain} specialist with ${years} years of experience.\nYour expertise includes: ${expertise_list}\nYou approach problems by: ${methodology}\n```\n\n### Pattern 2: Constraint-First Design\n```markdown\n# Constraints (Read First)\nMUST: ${required_constraints}\nMUST NOT: ${forbidden_actions}\nOPTIMIZE FOR: ${optimization_targets}\n\n# Task\n[task description with constraints in mind]\n```\n\n### Pattern 3: Format-Driven Output\n```markdown\nYour output MUST follow this exact structure:\n\n\\`\\`\\`json\n{\n  \"analysis\": \"...\",\n  \"recommendations\": [...],\n  \"confidence\": 0.XX\n}\n\\`\\`\\`\n\nDo not deviate from this format.\n```\n\n## Integration Examples\n\n### With Agent Creation\n```javascript\n\u002F\u002F Use prompt architect to optimize agent system prompts\nconst agentPrompt = await optimizePrompt({\n  role: 'backend-developer',\n  domain: 'Node.js API development',\n  constraints: ['RESTful design', 'security-first', 'test-driven'],\n  outputFormat: 'production-ready code'\n});\n```\n\n### With SPARC Workflow\n```bash\n# Optimize specification prompts\nnpx claude-flow@alpha skill-run prompt-architect \\\n  --input \"sparc-spec-prompt.md\" \\\n  --output \"optimized-spec-prompt.md\"\n\n# Use optimized prompt in SPARC\nnpx claude-flow@alpha sparc run spec-pseudocode \\\n  --prompt-file \"optimized-spec-prompt.md\"\n```\n\n## Memory Schema\n\n```javascript\n{\n  \"prompt-architect\u002F\": {\n    \"session-${id}\u002F\": {\n      \"analysis\": {},\n      \"anti-patterns\": [],\n      \"missing\": [],\n      \"structure\": {},\n      \"techniques-applied\": [],\n      \"test-cases\": [],\n      \"ab-test-results\": {},\n      \"metrics\": {},\n      \"failure-analysis\": [],\n      \"refined-prompt\": \"\",\n      \"final-prompt\": \"\"\n    }\n  }\n}\n```\n\n## Skill Completion\n\nOutputs:\n1. **final-prompt.md**: Optimized prompt ready for use\n2. **optimization-report.md**: Detailed analysis and results\n3. **ab-test-results.json**: Performance comparison data\n4. **prompt-library-entry.md**: Cataloged for reuse\n\nSkill complete when metrics show >20% improvement and all test cases pass.\n",{"name":5,"tags":52,"agents":56,"outputs":59,"version":63,"category":64,"triggers":65,"difficulty":70,"description":7,"dependencies":71,"prerequisites":73,"success_criteria":75,"validation_method":80,"estimated_duration":81},[39,53,54,55],"optimization","ai-systems","llm",[57,58],"researcher","coder",[60,61,62],"Optimized prompt","Analysis report","A\u002FB test results","1.0.0","utilities",[66,67,68,69],"Poor AI response quality","Inconsistent outputs","New prompt needed","Prompt refactoring required","intermediate",[72],"claude-flow@alpha",[74],"Prompt to optimize OR use case defined",[76,77,78,79],"Prompt structure optimized","Anti-patterns eliminated","Evidence-based techniques applied","Effectiveness validated","A\u002FB testing","20-40min","main","def8531e65114c0fca8fb8551c1871ee0eed705c","skills\u002Faiskillstore\u002Fmarketplace\u002Fdnyoussef\u002Fwhen-optimizing-prompts-use-prompt-architect",24,"https:\u002F\u002Fgithub.com\u002FComeOnOliver\u002Fskillshub","https:\u002F\u002Fskillshub.wtf","https:\u002F\u002Fgithub.com\u002FComeOnOliver","AI & Blockchain Engineer\r\n","2026-06-24 22:07:19",{"source":8,"skillName":5,"commands":92,"prompt":102,"zipAvailable":103},[93,96,99],{"tool":94,"command":95},"npx","npx skills add ComeOnOliver\u002Fskillshub --skill when-optimizing-prompts-use-prompt-architect",{"tool":97,"command":98},"bunx","bunx skills add ComeOnOliver\u002Fskillshub --skill when-optimizing-prompts-use-prompt-architect",{"tool":100,"command":101},"pnpm","pnpm dlx skills add ComeOnOliver\u002Fskillshub --skill when-optimizing-prompts-use-prompt-architect","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FComeOnOliver\u002Fskillshub\u002Fblob\u002Fmain\u002Fskills\u002Faiskillstore\u002Fmarketplace\u002Fdnyoussef\u002Fwhen-optimizing-prompts-use-prompt-architect\u002FSKILL.md",true,[105,111,117,121],{"path":106,"resourceType":107,"resourceTypeDesc":108,"sizeBytes":109,"sourceHtmlUrl":110},"process-diagram.gv","OTHER","其他",3548,"https:\u002F\u002Fgithub.com\u002FComeOnOliver\u002Fskillshub\u002Fblob\u002Fmain\u002Fskills\u002Faiskillstore\u002Fmarketplace\u002Fdnyoussef\u002Fwhen-optimizing-prompts-use-prompt-architect\u002Fprocess-diagram.gv",{"path":112,"resourceType":113,"resourceTypeDesc":114,"sizeBytes":115,"sourceHtmlUrl":116},"PROCESS.md","REFERENCE","参考文档",6832,"https:\u002F\u002Fgithub.com\u002FComeOnOliver\u002Fskillshub\u002Fblob\u002Fmain\u002Fskills\u002Faiskillstore\u002Fmarketplace\u002Fdnyoussef\u002Fwhen-optimizing-prompts-use-prompt-architect\u002FPROCESS.md",{"path":118,"resourceType":113,"resourceTypeDesc":114,"sizeBytes":119,"sourceHtmlUrl":120},"README.md",1556,"https:\u002F\u002Fgithub.com\u002FComeOnOliver\u002Fskillshub\u002Fblob\u002Fmain\u002Fskills\u002Faiskillstore\u002Fmarketplace\u002Fdnyoussef\u002Fwhen-optimizing-prompts-use-prompt-architect\u002FREADME.md",{"path":122,"resourceType":113,"resourceTypeDesc":114,"sizeBytes":123,"sourceHtmlUrl":124},"skill-report.json",24700,"https:\u002F\u002Fgithub.com\u002FComeOnOliver\u002Fskillshub\u002Fblob\u002Fmain\u002Fskills\u002Faiskillstore\u002Fmarketplace\u002Fdnyoussef\u002Fwhen-optimizing-prompts-use-prompt-architect\u002Fskill-report.json",[126,136,146,155,161,168],{"skillNo":127,"name":128,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":129,"categories":130,"keywords":132,"publishedAt":47},"01M6FKJN13GMG4KY57J3QD8GY5","model-hierarchy-skill","https:\u002F\u002Fgithub.com\u002FComeOnOliver\u002Fskillshub\u002Fblob\u002Fmain\u002Fskills\u002Fzscole\u002Fmodel-hierarchy-skill\u002Fmodel-hierarchy-skill\u002FSKILL.md",[131],{"categoryNo":24,"categoryName":25,"taxonomyType":21,"taxonomyTypeDesc":22},[133,134,135],"model","model hierarchy","model hierarchy skill",{"skillNo":137,"name":138,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":139,"categories":140,"keywords":142,"publishedAt":47},"01M6FKJJJ6GCAMR4PPPRD12M5X","zoom-mcp","https:\u002F\u002Fgithub.com\u002FComeOnOliver\u002Fskillshub\u002Fblob\u002Fmain\u002Fskills\u002Fzoom\u002Fskills\u002Fzoom-mcp\u002FSKILL.md",[141],{"categoryNo":24,"categoryName":25,"taxonomyType":21,"taxonomyTypeDesc":22},[143,144,145],"MCP","zoom","zoom mcp",{"skillNo":147,"name":148,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":149,"categories":150,"keywords":151,"publishedAt":47},"01M6FKJE3KYT3KPEZMEEHYKA6K","zoom-apps-sdk","https:\u002F\u002Fgithub.com\u002FComeOnOliver\u002Fskillshub\u002Fblob\u002Fmain\u002Fskills\u002Fzoom\u002Fskills\u002Fzoom-apps-sdk\u002FSKILL.md",[],[152,153,154,144],"sdk","zoom apps sdk","apps",{"skillNo":156,"name":157,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":158,"categories":159,"keywords":160,"publishedAt":47},"01M6FKJ9EDH9Q5VWDF8HAJRSP0","windows","https:\u002F\u002Fgithub.com\u002FComeOnOliver\u002Fskillshub\u002Fblob\u002Fmain\u002Fskills\u002Fzoom\u002Fskills\u002Fwindows\u002FSKILL.md",[],[157],{"skillNo":162,"name":163,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":164,"categories":165,"keywords":167,"publishedAt":47},"01M6FKJ666BEEFC3AE9GJ72Q93","whiteboard","https:\u002F\u002Fgithub.com\u002FComeOnOliver\u002Fskillshub\u002Fblob\u002Fmain\u002Fskills\u002Fzoom\u002Fskills\u002Fwhiteboard\u002FSKILL.md",[166],{"categoryNo":24,"categoryName":25,"taxonomyType":21,"taxonomyTypeDesc":22},[163],{"skillNo":169,"name":170,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":171,"categories":172,"keywords":173,"publishedAt":47},"01M6FKJ21WFCDGTJQ2Q91T9DVW","websockets","https:\u002F\u002Fgithub.com\u002FComeOnOliver\u002Fskillshub\u002Fblob\u002Fmain\u002Fskills\u002Fzoom\u002Fskills\u002Fwebsockets\u002FSKILL.md",[],[170]]