When to Use This Skill

Use this skill when:

• Code quality issues are detected (violations, smells, anti-patterns)
• Audit requirements mandate systematic review (compliance, release gates)
• Review needs arise (pre-merge, production hardening, refactoring preparation)
• Quality metrics indicate degradation (test coverage drop, complexity increase)
• Theater detection is needed (mock data, stubs, incomplete implementations)

When NOT to Use This Skill

Do NOT use this skill for:

• Simple formatting fixes (use linter/prettier directly)
• Non-code files (documentation, configuration without logic)
• Trivial changes (typo fixes, comment updates)
• Generated code (build artifacts, vendor dependencies)
• Third-party libraries (focus on application code)

Success Criteria

This skill succeeds when:

• Violations Detected: All quality issues found with ZERO false negatives
• False Positive Rate: <5% (95%+ findings are genuine issues)
• Actionable Feedback: Every finding includes file path, line number, and fix guidance
• Root Cause Identified: Issues traced to underlying causes, not just symptoms
• Fix Verification: Proposed fixes validated against codebase constraints

Edge Cases and Limitations

Handle these edge cases carefully:

• Empty Files: May trigger false positives - verify intent (stub vs intentional)
• Generated Code: Skip or flag as low priority (auto-generated files)
• Third-Party Libraries: Exclude from analysis (vendor/, node_modules/)
• Domain-Specific Patterns: What looks like violation may be intentional (DSLs)
• Legacy Code: Balance ideal standards with pragmatic technical debt management

Quality Analysis Guardrails

CRITICAL RULES - ALWAYS FOLLOW:

• NEVER approve code without evidence: Require actual execution, not assumptions
• ALWAYS provide line numbers: Every finding MUST include file:line reference
• VALIDATE findings against multiple perspectives: Cross-check with complementary tools
• DISTINGUISH symptoms from root causes: Report underlying issues, not just manifestations
• AVOID false confidence: Flag uncertain findings as "needs manual review"
• PRESERVE context: Show surrounding code (5 lines before/after minimum)
• TRACK false positives: Learn from mistakes to improve detection accuracy

Evidence-Based Validation

Use multiple validation perspectives:

1. Static Analysis: Code structure, patterns, metrics (connascence, complexity)
2. Dynamic Analysis: Execution behavior, test results, runtime characteristics
3. Historical Analysis: Git history, past bug patterns, change frequency
4. Peer Review: Cross-validation with other quality skills (functionality-audit, theater-detection)
5. Domain Expertise: Leverage .claude/expertise/{domain}.yaml if available

Validation Threshold: Findings require 2+ confirming signals before flagging as violations.

Integration with Quality Pipeline

This skill integrates with:

• Pre-Phase: Load domain expertise (.claude/expertise/{domain}.yaml)
• Parallel Skills: functionality-audit, theater-detection-audit, style-audit
• Post-Phase: Store findings in Memory MCP with WHO/WHEN/PROJECT/WHY tags
• Feedback Loop: Learnings feed dogfooding-system for continuous improvement

Dogfooding System - Self-Improvement Architecture

A comprehensive 3-phase self-improvement system that enables Claude Code to automatically improve itself and connected MCP servers through quality detection, pattern retrieval, and safe automated fixes.

Overview

The Dogfooding System orchestrates three integrated phases:

1. Quality Detection - Run Connascence Analysis, detect 7+ violation types, store in Memory-MCP
2. Pattern Retrieval - Query Memory-MCP via vector search, rank patterns by similarity + success rate
3. Continuous Improvement - Full cycle orchestration with sandbox testing and automated rollback

Key Components:

• Claude Code: 106+ skills, 131+ agents for orchestration
• Connascence Analyzer MCP: 7+ violation types (God Objects, Parameter Bombs, Deep Nesting, etc.)
• Memory-MCP Triple System: Persistent cross-session memory with semantic search and metadata tagging

When to Use This Skill

Activate this skill when:

• Need to analyze code quality across projects systematically
• Want to find and apply past fix patterns for similar violations
• Running automated improvement cycles on MCP servers or Claude Code itself
• Implementing self-healing workflows with safety guarantees
• Building institutional knowledge about code quality patterns
• Establishing continuous improvement processes

This skill is particularly valuable when:

• Working with large codebases requiring systematic quality improvement
• Building self-improving systems that learn from past fixes
• Implementing production-ready dogfooding workflows with safety rails
• Creating automated code quality pipelines with verification

Three-Phase Architecture

Phase 1: Quality Detection (30-60 seconds)

Run Connascence Analysis to detect violations and store findings in Memory-MCP.

Workflow:

1. Run Connascence Analyzer on target project
2. Parse 7+ violation types with severity levels
3. Store each violation in Memory-MCP with WHO/WHEN/PROJECT/WHY tags
4. Update Grafana dashboard with metrics
5. Generate summary report

Violations Detected:

• God Objects (>15 methods)
• Parameter Bombs/CoP (>6 params, NASA limit)
• Cyclomatic Complexity (>10)
• Deep Nesting (>4 levels, NASA limit)
• Long Functions (>50 lines)
• Magic Literals/CoM (hardcoded values)
• Duplicate Code Blocks

Execution:

# Run quality detection for single project
.\resources\scripts\run-quality-detection.bat memory-mcp

# Run for all projects
.\resources\scripts\run-quality-detection.bat all

Agents: code-analyzer, reviewer

Outputs:

• JSON: metrics/dogfooding/<project>_<timestamp>.json
• Summary: metrics/dogfooding/summary_<timestamp>.txt
• Memory-MCP: Tagged violation records
• Dashboard: Updated metrics

Phase 2: Pattern Retrieval (10-30 seconds)

Query Memory-MCP for similar past fixes using vector search, rank by relevance, optionally apply.

Workflow:

1. Receive violation from Phase 1 (or manual query)
2. Generate 384-dimensional embedding via all-MiniLM-L6-v2
3. Query ChromaDB with HNSW indexing for top-5 similar patterns
4. Rank by: similarity (40%) + success_rate (30%) + context_match (20%) + recency (10%)
5. Select best transformation strategy
6. (Optional) Apply AST transformation with sandbox testing

Vector Search:

• Model: all-MiniLM-L6-v2 (384-dimensional embeddings)
• Backend: ChromaDB with HNSW indexing
• Similarity: Cosine distance (1 - distance = similarity score)
• Filters: intent, project, agent_category metadata

Transformation Strategies:

1. Delegation Pattern - God Object → separate classes
2. Config Object Pattern - Parameter Bomb → object param
3. Early Return Pattern - Deep Nesting → guard clauses
4. Extract Method Pattern - Long Function → smaller functions
5. Named Constant Pattern - Magic Literal → const
6. Extract Function Pattern - Duplicate Code → DRY

Execution:

# Query only (no application)
.\resources\scripts\run-pattern-retrieval.bat "God Object with 26 methods"

# Query + apply best pattern
.\resources\scripts\run-pattern-retrieval.bat "Parameter Bomb 10 params" --apply

Agents: code-analyzer, coder, reviewer

Outputs:

• JSON: retrievals/query-<timestamp>.json (vector search results)
• Best pattern: retrievals/best-pattern-<timestamp>.json
• (Optional) Applied transformation with git commit

Phase 3: Continuous Improvement (60-120 seconds)

Full cycle orchestration combining Quality Detection + Pattern Retrieval + Safe Application.

Workflow:

1. Quality Detection (30-60s) - Detect violations
2. Pattern Retrieval (10-30s) - Find best fixes
3. Safe Application (20-40s) - Apply in sandbox first
4. Verification (15s) - Re-run analysis, compare metrics
5. Summary & Metrics (10-20s) - Generate reports, update dashboard

Safety Checks (MANDATORY):

1. ✅ Sandbox testing REQUIRED before production
2. ✅ Automated rollback via git stash
3. ✅ Progressive application (one fix at a time)
4. ✅ Test coverage ≥70% required
5. ✅ CI/CD gate must pass before merge

Execution:

# Single cycle with safety checks
.\resources\scripts\run-continuous-improvement.bat memory-mcp

# Dry-run (no fixes applied)
.\resources\scripts\run-continuous-improvement.bat memory-mcp --dry-run

# Full cycle all projects (round-robin)
.\resources\scripts\run-continuous-improvement.bat all

Agents: hierarchical-coordinator, code-analyzer, coder, reviewer

Outputs:

• Cycle summary: cycle-summaries/cycle-<id>.txt
• Archive: archive/<cycle_id>/ (all artifacts)
• Dashboard: Updated Grafana metrics
• Memory-MCP: Cycle summary + all fixes stored
• Git commits: Fixes with safety metadata

Memory-MCP Tagging Protocol

ALL Memory-MCP writes use automatic metadata tagging:

WHO: Agent name, category, capabilities WHEN: ISO timestamp, Unix timestamp, readable format PROJECT: connascence-analyzer, memory-mcp-triple-system, claude-flow, etc. WHY: Intent (implementation, bugfix, refactor, testing, documentation, analysis, planning, research)

Implementation:

const { taggedMemoryStore } = require('./hooks/12fa/memory-mcp-tagging-protocol.js');

// Auto-tagged memory write
const tagged = taggedMemoryStore('code-analyzer', 'Detected God Object with 26 methods', {
  violation_type: 'god-object',
  severity: 'high',
  project: 'memory-mcp'
});
// Automatically includes: agent metadata, timestamps, project, intent

Safety Rules (CRITICAL)

Documentation: C:\Users\17175\docs\DOGFOODING-SAFETY-RULES.md

1. Sandbox Testing REQUIRED

mkdir C:\Users\17175\tmp\dogfood-sandbox
xcopy /E /I /Q <project> C:\Users\17175\tmp\dogfood-sandbox
cd C:\Users\17175\tmp\dogfood-sandbox && npm test
# If pass → apply to production
# If fail → reject fix

2. Automated Rollback

git stash push -u -m "backup-<timestamp>"
<apply-fix>
npm test || git stash pop  # Rollback on failure

3. Progressive Application

• Fix ONE violation at a time
• Test after each fix
• Commit after each successful fix
• Never batch fixes without testing

4. Test Coverage Requirement

• ONLY apply fixes to code with ≥70% test coverage
• Add tests FIRST if coverage is insufficient

5. CI/CD Gate

• ALL fixes must pass CI/CD before merge
• Automated safety checks in .github/workflows/dogfooding-safety.yml

Metrics Tracked

Phase 1 (Quality Detection)

• Analysis Duration (target: <30s)
• Violations Detected (by type)
• Files Analyzed
• Storage Success Rate (target: 100%)
• Dashboard Update Status

Phase 2 (Pattern Retrieval)

• Query Time (target: <1000ms)
• Results Found (target: ≥3)
• Top Similarity Score (target: ≥0.70)
• Application Success Rate (target: ≥95%)
• Improvement Magnitude

Phase 3 (Continuous Improvement)

• Cycle Duration (target: <120s)
• Violations Fixed (target: ≥3 per cycle)
• Success Rate (target: ≥95%)
• Improvement Velocity (target: ≥5 violations/day)
• Pattern Retrieval Quality (target: ≥0.75 avg similarity)
• Sandbox Testing Pass Rate (target: 100%)
• Production Rollback Rate (target: ≤5%)

Quick Start

1. Run Quality Detection Only

.\resources\scripts\run-quality-detection.bat memory-mcp

2. Query Past Fixes

.\resources\scripts\run-pattern-retrieval.bat "God Object with 26 methods"

3. Run Full Improvement Cycle

.\resources\scripts\run-continuous-improvement.bat memory-mcp

4. Schedule Automated Cycles

# Windows Task Scheduler (daily at 12:00 UTC)
schtasks /create /tn "Dogfooding-Cycle" \
  /tr "C:\Users\17175\claude-code-plugins\ruv-sparc-three-loop-system\skills\dogfooding-system\resources\scripts\run-continuous-improvement.bat all" \
  /sc daily /st 12:00

Bundled Resources

Scripts

• run-quality-detection.bat - Execute Phase 1 (Quality Detection)
• run-pattern-retrieval.bat - Execute Phase 2 (Pattern Retrieval)
• run-continuous-improvement.bat - Execute Phase 3 (Full Cycle)
• generate-cycle-summary.js - Generate cycle summary reports

Templates

• violation-report.md - Violation report template with metadata
• fix-pattern.json - Fix pattern schema for Memory-MCP storage
• cycle-summary.md - Cycle summary template with metrics

Tests

• test-quality-detection.js - Verify Phase 1 workflow
• test-pattern-retrieval.js - Verify Phase 2 vector search
• test-continuous-improvement.js - Verify Phase 3 full cycle

Troubleshooting

Issue: VectorIndexer has no attribute 'collection'

Fix: Already patched in Memory-MCP

def __init__(self, ...):
    self.client = chromadb.PersistentClient(path=persist_directory)
    self.create_collection()  # <-- Added this line

Issue: No patterns found in Memory-MCP

Solution: Run Phase 1 (Quality Detection) first to populate Memory-MCP with violations and fixes

Issue: Sandbox tests pass but production fails

Solution: Enhanced sandbox testing to better replicate production environment

• Check for environment-specific dependencies
• Verify test coverage is sufficient
• Update sandbox creation to match production more closely

Integration with Other Skills

Upstream Skills (that feed into dogfooding):

• functionality-audit - Sandbox testing + debugging
• theater-detection-audit - Byzantine consensus verification
• production-readiness - Complete deployment checklist

Downstream Skills (that use dogfooding):

• cicd-intelligent-recovery - Automated failure recovery
• performance-analysis - Performance optimization
• code-review-assistant - Multi-agent swarm review

File Structure

dogfooding-system/
├── skill.md (this file)
├── INDEX.md (comprehensive documentation)
├── resources/
│   ├── scripts/
│   │   ├── run-quality-detection.bat
│   │   ├── run-pattern-retrieval.bat
│   │   ├── run-continuous-improvement.bat
│   │   └── generate-cycle-summary.js
│   └── templates/
│       ├── violation-report.md
│       ├── fix-pattern.json
│       └── cycle-summary.md
├── tests/
│   ├── test-quality-detection.js
│   ├── test-pattern-retrieval.js
│   └── test-continuous-improvement.js
└── graphviz/
    └── dogfooding-system-process.dot

Success Metrics

Since implementation:

• ✅ VectorIndexer bug fixed (collection attribute initialization)
• ✅ 27 Unicode violations fixed in Connascence Analyzer
• ✅ 7 import issues fixed in Memory-MCP
• ✅ 45 violations detected in Memory-MCP codebase
• ✅ 46 fixes stored in Memory-MCP with proper metadata
• ✅ Vector search working with 0.82+ average similarity
• ✅ WHO/WHEN/PROJECT/WHY tagging protocol implemented
• ✅ Safety rules documented and enforced
• ✅ 3 SOP skills created with full agent assignments

Status: ✅ PRODUCTION READY Version: 1.0 (Gold Tier) Skills: 3 SOPs (Quality Detection, Pattern Retrieval, Continuous Improvement) Agents: hierarchical-coordinator, code-analyzer, coder, reviewer MCP Tools: connascence-analyzer, memory-mcp, claude-flow Safety: Sandbox testing + automated rollback + verification

Core Principles

1. Evidence-Based Detection Over Assumptions

Quality violations must be detected through measurable metrics (connascence analysis), not subjective judgment. Every finding requires concrete evidence.

In practice:

• Use Connascence Analyzer for 7+ violation types with thresholds
• Require 2+ confirming signals before flagging violations
• Store findings with WHO/WHEN/PROJECT/WHY metadata
• Track false positive rates (<5% target)

2. Progressive Improvement Through Pattern Learning

Fix quality issues by learning from past solutions, not reinventing approaches. Build institutional knowledge through vector search.

In practice:

• Store every fix in Memory-MCP with context
• Query similar violations before applying new fixes
• Rank patterns by similarity (40%) + success rate (30%) + context (20%) + recency (10%)
• Update patterns after successful fixes

3. Safety-First Automated Fixes

Apply fixes in isolated sandboxes with automated rollback before touching production code.

In practice:

• Test EVERY fix in sandbox before production
• Use git stash for automatic rollback on failure
• Fix ONE violation at a time with validation
• Require test coverage >=70% before applying fixes

Anti-Patterns

Anti-Pattern	Problem	Solution
Batch Fixes Without Testing	Applying multiple fixes simultaneously causes cascading failures with unknown root cause	Fix ONE violation at a time, test after each
Skipping Sandbox Validation	Direct production fixes risk breaking functionality with no recovery path	ALWAYS test in sandbox first, rollback on failure
Ignoring Pattern History	Reinventing solutions wastes time and introduces untested approaches	Query Memory-MCP for similar violations before fixing
Low Confidence Threshold	Flagging ambiguous patterns creates false positives and noise	Use >=40 confidence score threshold, require 2+ signals
Missing Metadata Tagging	Fixes stored without WHO/WHEN/PROJECT/WHY context become unusable for pattern retrieval	Use taggedMemoryStore() for automatic metadata injection

Conclusion

The Dogfooding System represents a paradigm shift from reactive debugging to proactive quality improvement through institutional learning. By combining Connascence Analysis for detection, Memory-MCP for pattern storage, and sandbox testing for safe application, organizations build self-improving codebases that learn from past fixes.

The three-phase architecture (Quality Detection -> Pattern Retrieval -> Continuous Improvement) creates a feedback loop where every fix strengthens future detection. Teams implementing dogfooding should prioritize safety rules (sandbox testing, progressive fixes, test coverage requirements) and proper metadata tagging for pattern retrieval. The system's vector search capabilities enable finding relevant fixes even when violation details differ, as semantic similarity identifies underlying patterns.

Most critically, dogfooding succeeds when organizations resist the temptation to batch fixes or skip validation. The discipline of one-fix-at-a-time testing, combined with automated rollback, creates sustainable quality improvement without production risk. Teams measuring success by improvement velocity (violations fixed per day) rather than just violation counts build codebases that continuously evolve toward higher quality.