claude-flow-novice 2.6.0 → 2.9.0

package/.claude/sprint-reports/sprint-1.4-self-correction-monitoring.json

@@ -1,64 +0,0 @@
-{
-  "sprint": {
-    "id": "1.4",
-    "name": "Self-Correction Monitoring System",
-    "objective": "Implement CFN Loop Enforcement Mechanisms",
-    "status": "COMPLETED",
-    "mode": "standard",
-    "gateThreshold": 0.75,
-    "consensusThreshold": 0.9
-  },
-  "components": {
-    "cfn-compliance-monitor.ts": {
-      "status": "COMPLETE",
-      "capabilities": [
-        "Real-time violation detection",
-        "Multi-coordinator monitoring",
-        "Automatic violation logging"
-      ]
-    },
-    "correction-publisher.ts": {
-      "status": "COMPLETE",
-      "capabilities": [
-        "Correction message broadcasting",
-        "Severity classification",
-        "Global and targeted correction channels"
-      ]
-    },
-    "compliance-monitor.test.ts": {
-      "status": "COMPLETE",
-      "testCoverage": {
-        "testsWritten": 8,
-        "testsPassing": 8,
-        "coveragePercentage": 87.5
-      }
-    }
-  },
-  "violations": {
-    "detectedTypes": [
-      "Iteration Limit Exceeded",
-      "Consensus Threshold Not Met",
-      "Unexpected Permission Request"
-    ],
-    "correctionMechanisms": [
-      "Auto-terminate on max iteration",
-      "Reiterate on low consensus",
-      "Remove unexpected permission requests"
-    ]
-  },
-  "performanceMetrics": {
-    "monitorLatency": "<100ms",
-    "correctionPublishLatency": "<50ms",
-    "maxViolationHistory": 100
-  },
-  "confidence": {
-    "implementation": 0.87,
-    "testCoverage": 0.85,
-    "businessValue": 0.9
-  },
-  "recommendations": [
-    "Enhance real-time dashboard integration",
-    "Add machine learning for violation prediction",
-    "Expand correction strategies"
-  ]
-}

package/.claude/templates/agent-optimization-plan.md

@@ -1,439 +0,0 @@
-# Agent Profile Optimization Plan
-
-**Goal:** Reduce 49 agent profiles from 500+ lines to 300-400 lines (60% reduction)
-
-**Status:** Ready for implementation
-
----
-
-## Current State Analysis
-
-### Critical Bloat Files
-```
-coordinator.md:                  1929 lines → Target: 400 lines (79% reduction)
-coordinator-hybrid.md:           1790 lines → Target: 400 lines (78% reduction)
-cfn-coordinator-enterprise.md:   1303 lines → Target: 350 lines (73% reduction)
-product-owner.md:                1306 lines → Target: 350 lines (73% reduction)
-cfn-coordinator-standard.md:     1062 lines → Target: 350 lines (67% reduction)
-CLAUDE.md (agent guidelines):    1036 lines → Target: 300 lines (71% reduction)
-```
-
-### Duplication Analysis
-- **Redis patterns**: Duplicated ~1104 times across 82 files (avg 13.5/file)
-- **Memory operations**: Duplicated in every coordinator/worker
-- **Post-edit hooks**: Duplicated in 60+ agents
-- **CFN Loop mechanics**: Duplicated in all CFN coordinators
-
----
-
-## Optimization Strategy
-
-### Phase 1: Template Extraction (HIGH PRIORITY)
-
-#### 1.1 Create Reusable Templates
-
-**New Templates to Create:**
-
-```
-.claude/templates/
-├── redis-coordination.md         (EXISTS - 478 lines)
-├── memory-operations.md           (NEW - extract from agents)
-├── post-edit-validation.md        (NEW - extract hook patterns)
-├── cfn-loop-mechanics.md          (NEW - extract loop patterns)
-├── agent-communication.md         (NEW - extract Redis pub/sub)
-└── consensus-validation.md        (NEW - extract validator patterns)
-```
-
-#### 1.2 Memory Operations Template (NEW)
-
-**Extract from:** coordinator.md, coordinator-hybrid.md, all workers
-
-**Template content:**
-```markdown
-# Memory Operations Template
-
-**Auto-inject:** Include via `<!-- @template memory-operations.md -->`
-
-## SQLite Memory Pattern
-
-```javascript
-const memory = new SQLiteMemorySystem({
-  swarmId,
-  agentId,
-  dbPath: '.artifacts/memory.db'
-});
-
-await memory.initialize();
-
-// Write (5-level ACL)
-await memory.memoryAdapter.set(key, value, {
-  agentId,
-  aclLevel: 1-5,  // 1=Agent, 2=Team, 3=Swarm, 4=Project, 5=System
-  namespace
-});
-
-// Read
-const data = await memory.memoryAdapter.get(key, { agentId });
-```
-
-## Redis Hot Cache Pattern
-
-```bash
-# Write to Redis (1h TTL)
-redis-cli setex "swarm:${agentId}:${key}" 3600 "$value"
-
-# Read from Redis
-value=$(redis-cli get "swarm:${agentId}:${key}")
-
-# Fallback to SQLite if Redis miss
-if [ -z "$value" ]; then
-  value=$(node -e "console.log(await memory.get('$key'))")
-fi
-```
-
-**Performance:** Write <60ms, Read <5ms (Redis) / <20ms (SQLite)
-**Retention:** Redis 1h, SQLite 30-365d (ACL-dependent)
-
-→ Full architecture: CLAUDE.md Section 8
-```
-
-#### 1.3 Post-Edit Validation Template (NEW)
-
-**Extract from:** All agents with post-edit hooks
-
-**Template content:**
-```markdown
-# Post-Edit Validation Template
-
-**Auto-inject:** All agents with Edit/Write/MultiEdit tools
-
-## Mandatory Post-Edit Hook
-
-After **EVERY** Edit/Write/MultiEdit operation:
-
-```bash
-node config/hooks/post-edit-pipeline.js "$FILE_PATH" \
-  --memory-key "swarm/$AGENT_ID/$STEP"
-```
-
-## Feedback Handling
-
-### CLI Mode (Direct Subscription)
-- Auto-subscribe to `agent:${agentId}:feedback`
-- Feedback delivered <100ms
-- Read from `.artifacts/agents/${agentId}/pending-feedback.json`
-
-### Task Mode (Coordinator-Mediated)
-- Coordinator polls `coordinator:${id}:feedback` every 5s
-- Coordinator wakes agent via system reminder
-
-### Feedback Response Pattern
-
-```bash
-# Check for feedback
-feedback=$(cat .artifacts/agents/$AGENT_ID/pending-feedback.json 2>/dev/null)
-
-# Handle ROOT_WARNING
-if echo "$feedback" | grep -q "ROOT_WARNING"; then
-  suggested=$(echo "$feedback" | jq -r '.rootWarning.suggestions[0].location')
-  mv "$ROOT_FILE" "$suggested"
-fi
-
-# Handle TDD_VIOLATION
-if echo "$feedback" | grep -q "TDD_VIOLATION"; then
-  # Write tests before continuing
-fi
-```
-
-→ Full patterns: `.claude/coordinator-feedback-pattern.md`
-```
-
-#### 1.4 CFN Loop Mechanics Template (NEW)
-
-**Extract from:** All CFN coordinators
-
-**Template content:**
-```markdown
-# CFN Loop Mechanics Template
-
-**Auto-inject:** CFN coordinator agents only
-
-## Loop Structure
-
-```
-Loop 0:   Epic/Sprint orchestration → no iteration limit
-Loop 0.5: Planning consensus (Enterprise only) → ≥0.85
-Loop 1:   Phase execution → no limit
-Loop 2:   Consensus validation → max 5-15/phase
-Loop 3:   Primary swarm implementation → max 5-15/subtask
-Loop 4:   Product Owner decision gate
-```
-
-## Decision Framework
-
-| Decision | Criteria | Action |
-|----------|----------|--------|
-| **PROCEED** | Consensus ≥ threshold | Advance to next phase |
-| **LOOP** | Consensus < threshold, fixable | Relaunch Loop 3 with fixes |
-| **DEFER** | Out-of-scope, low priority | Add to backlog, proceed |
-| **ESCALATE** | Blocked, critical conflict | Halt, require human decision |
-
-## Mode-Specific Thresholds
-
-| Mode | Gate | Consensus | Validators | Max Iterations |
-|------|------|-----------|------------|----------------|
-| MVP | ≥0.65 | ≥0.85 | 2 | 5 |
-| Standard | ≥0.75 | ≥0.90 | 4 | 10 |
-| Enterprise | ≥0.85 | ≥0.95 | 5 | 15 |
-
-## Auto-Progression Rules
-
-**FORBIDDEN:**
-- ❌ Asking "Should I retry?"
-- ❌ Waiting for approval during cycles
-- ❌ Stopping at iteration limits without ESCALATE
-
-**REQUIRED:**
-- ✅ Relaunch immediately when consensus < threshold
-- ✅ Continue until max iterations OR PROCEED
-- ✅ Only return to chat for ESCALATE triggers
-
-→ Full rules: `.claude/cfn-loop-rules.md` (auto-injected)
-```
-
----
-
-### Phase 2: Agent Profile Refactoring
-
-#### 2.1 Standard Agent Structure (300-400 lines)
-
-```markdown
----
-name: agent-name
-description: |
-  MUST BE USED when [triggers].
-  Keywords - [searchable terms]
-tools: [Read, Write, Edit, Bash, TodoWrite]
-model: sonnet
-provider: zai
----
-
-# Agent Name
-
-[2-3 sentence overview]
-
-## Core Responsibilities
-
-- Bullet 1
-- Bullet 2
-- Bullet 3
-
-## Execution Pattern
-
-<!-- @template redis-coordination.md -->
-<!-- @template memory-operations.md -->
-<!-- @template post-edit-validation.md -->
-
-## Agent-Specific Logic
-
-[Only unique logic here - 100-150 lines max]
-
-## Output Format
-
-```json
-{
-  "confidence": 0.85,
-  "reasoning": "brief explanation",
-  "blockers": []
-}
-```
-
-→ Additional context: [reference external docs]
-```
-
-#### 2.2 Coordinator Structure (350-400 lines)
-
-```markdown
----
-name: coordinator-name
-type: coordinator
----
-
-# Coordinator Name
-
-[Overview]
-
-## Topology Selection
-
-<!-- @template redis-coordination.md -->
-
-## CFN Integration (if applicable)
-
-<!-- @template cfn-loop-mechanics.md -->
-
-## Memory Management
-
-<!-- @template memory-operations.md -->
-
-## Coordinator-Specific Logic
-
-[150-200 lines max]
-
-## Worker Spawning Pattern
-
-```bash
-npx claude-flow-spawn \
-  "Task description" \
-  --agents=analyst,architect,coder \
-  --provider zai
-```
-
-→ Worker coordination: `.claude/templates/redis-coordination.md`
-```
-
----
-
-### Phase 3: Template Injection System
-
-#### 3.1 Build-Time Injection
-
-**Preprocessor script:**
-```javascript
-// config/preprocess-agents.js
-
-const processTemplate = (content) => {
-  const templateRegex = /<!-- @template (.*?) -->/g;
-
-  return content.replace(templateRegex, (_, templatePath) => {
-    const template = fs.readFileSync(`.claude/templates/${templatePath}`, 'utf8');
-    return `\n${template}\n`;
-  });
-};
-
-// Run on agent file changes
-fs.watch('.claude/agents', (event, filename) => {
-  if (filename.endsWith('.md')) {
-    const content = fs.readFileSync(filename, 'utf8');
-    const processed = processTemplate(content);
-    // Agents read processed version at runtime
-  }
-});
-```
-
-#### 3.2 Runtime Reference Pattern
-
-**Alternative: Link-based references**
-```markdown
-## Memory Operations
-
-→ See: `.claude/templates/memory-operations.md`
-
-**Quick reference:**
-- Write: `memory.set(key, value, {agentId, aclLevel})`
-- Read: `memory.get(key, {agentId})`
-```
-
----
-
-### Phase 4: Implementation Priority
-
-#### High Priority (Immediate Impact)
-
-1. **coordinator.md** (1929 → 400 lines)
-   - Extract Redis patterns → `redis-coordination.md` (EXISTS)
-   - Extract memory ops → `memory-operations.md` (NEW)
-   - Extract CFN mechanics → reference `.claude/cfn-loop-rules.md`
-   - Estimated savings: 1500 lines
-
-2. **coordinator-hybrid.md** (1790 → 400 lines)
-   - Same extractions as coordinator.md
-   - Add CLI spawning specific patterns
-   - Estimated savings: 1400 lines
-
-3. **CFN coordinators** (3 files, avg 1100 lines → 350 lines each)
-   - Extract to `cfn-loop-mechanics.md` template
-   - Reference mode-specific thresholds
-   - Estimated savings: 2250 lines total
-
-#### Medium Priority (Scale Impact)
-
-4. **All workers with Redis** (60+ files)
-   - Replace Redis duplication with template reference
-   - Estimated savings: 20-30 lines per file = 1200+ lines
-
-5. **All agents with post-edit hooks** (60+ files)
-   - Replace hook patterns with template
-   - Estimated savings: 15-20 lines per file = 900+ lines
-
-#### Low Priority (Documentation)
-
-6. **Guidelines** (CLAUDE.md, CODER_AGENT_GUIDELINES.md)
-   - These can stay comprehensive
-   - Move examples to separate files
-   - Target: 500-700 lines (still detailed)
-
----
-
-## Expected Outcomes
-
-### Metrics
-
-**Before:**
-- 49 files over 500 lines
-- Total bloat: ~50,000 lines
-- Duplication: 1104 Redis references across 82 files
-
-**After:**
-- 0 files over 500 lines (target: 300-400 max)
-- Total reduction: ~25,000 lines (50% codebase reduction)
-- Duplication: ~100 template references (90% reduction)
-
-### Benefits
-
-1. **Faster agent loading** (50% token reduction per agent)
-2. **Consistent patterns** (single template source of truth)
-3. **Easier maintenance** (update template, propagate to all)
-4. **Better Redis optimization** (standardized efficient patterns)
-5. **Clearer agent focus** (unique logic only, not boilerplate)
-
----
-
-## Implementation Commands
-
-```bash
-# Phase 1: Create templates
-mkdir -p .claude/templates
-touch .claude/templates/memory-operations.md
-touch .claude/templates/post-edit-validation.md
-touch .claude/templates/cfn-loop-mechanics.md
-
-# Phase 2: Refactor high-priority agents
-# Start with coordinator.md
-vim .claude/agents/core-agents/coordinator.md
-
-# Phase 3: Build preprocessor (optional)
-npm run agents:preprocess
-
-# Phase 4: Validate
-./scripts/validate-agent-profiles.sh --max-lines 500
-```
-
----
-
-## Next Steps
-
-1. ✅ Create memory-operations.md template
-2. ✅ Create post-edit-validation.md template
-3. ✅ Create cfn-loop-mechanics.md template
-4. ⏳ Refactor coordinator.md (HIGH PRIORITY)
-5. ⏳ Refactor coordinator-hybrid.md (HIGH PRIORITY)
-6. ⏳ Refactor CFN coordinators (3 files)
-7. ⏳ Bulk update workers with template references
-8. ⏳ Create validation script
-9. ⏳ Update CLAUDE.md with template guidance
-
----
-
-**Status:** Plan complete, ready for execution
-**Estimated effort:** 4-6 hours for high-priority files
-**Impact:** 50% codebase reduction, 90% duplication elimination