fix: Fungus MCP connection, task consolidation (465→~60), and EventBus-SoM bridge expansion

Three critical pipeline issues resolved:

1. Fungus MCP: Rewrote server from raw JSON-RPC to official MCP Python SDK,
   fixed config path (was writing to ~/.kilocode/ instead of .kilo/mcp.json),
   now shows "fungus: connected" in kilo mcp list.

2. Task Splitting: Added feature-consolidated frontend tasks (1 per component
   instead of 5) and test tasks (max 3 per epic instead of 1 per Gherkin
   scenario). Reduces ~465 tasks to ~50-60 in feature consolidation mode.

3. EventBus-SoM Bridge: Added 6 new MCP Event Bridge mappings
   (EPIC_TASK_COMPLETED/FAILED, API_ROUTES_GENERATED, AUTH_SETUP_COMPLETE,
   TYPE_ERROR) and 12 new SoM Bridge prefix mappings for feature-consolidated
   task types (verify_all, api_controller, fe_page, test_e2e_*, env_, etc).

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Author: Flissel

.claude/launch.json

@@ -0,0 +1,12 @@
+{
+  "version": "0.0.1",
+  "configurations": [
+    {
+      "name": "web-app-frontend",
+      "runtimeExecutable": "node",
+      "runtimeArgs": ["node_modules/vite/bin/vite.js", "dev", "--host", "--port", "8081"],
+      "port": 8081,
+      "cwd": "web-app/front"
+    }
+  ]
+}

.claude/skills/ce-code-quality/SKILL.md

@@ -0,0 +1,109 @@
+---
+name: ce-code-quality
+description: Review quality of generated code. Checks for empty files, broken imports, missing modules, duplicate code, TypeScript errors, and rates overall quality. Identifies files that need regeneration.
+trigger: When user asks to "review code quality", "check generated code", "is the code real", "any empty files", "code review the output"
+---
+
+# ce-code-quality — Generated Code Quality Audit
+
+## Step 1: File Inventory
+
+```bash
+docker exec coding-engine-api bash -c "
+cd /app/output/*/
+echo '=== FILE COUNTS ==='
+echo 'Controllers:' && find src/modules -name '*.controller.ts' 2>/dev/null | wc -l
+echo 'Services:' && find src/modules -name '*.service.ts' 2>/dev/null | wc -l
+echo 'DTOs:' && find src -name '*.dto.ts' 2>/dev/null | wc -l
+echo 'Guards:' && find src -name '*.guard.ts' 2>/dev/null | wc -l
+echo 'Validators:' && find src -name '*.validator.ts' 2>/dev/null | wc -l
+echo 'Frontend Pages:' && find frontend/src -name '*.tsx' 2>/dev/null | grep -v node_modules | wc -l
+echo 'Hooks:' && find src/hooks -name '*.ts' 2>/dev/null | wc -l
+"
+```
+
+## Step 2: Empty File Detection
+
+```bash
+docker exec coding-engine-api bash -c "
+cd /app/output/*/
+echo '=== EMPTY OR TINY FILES (< 50 bytes) ==='
+find src -type f -name '*.ts' -o -name '*.tsx' 2>/dev/null | grep -v node_modules | while read f; do
+  size=\$(wc -c < \"\$f\")
+  if [ \$size -lt 50 ]; then echo \"EMPTY: \$f (\${size}b)\"; fi
+done
+echo '=== PLACEHOLDER FILES (contain only imports/exports) ==='
+find src -type f -name '*.ts' 2>/dev/null | grep -v node_modules | while read f; do
+  lines=\$(wc -l < \"\$f\")
+  if [ \$lines -lt 5 ]; then echo \"STUB: \$f (\${lines} lines)\"; fi
+done | head -20
+"
+```
+
+## Step 3: Import Health Check
+
+```bash
+docker exec coding-engine-api bash -c "
+cd /app/output/*/
+echo '=== BROKEN IMPORTS ==='
+grep -rn 'from.*react-native' src/modules/ src/api/ 2>/dev/null | head -5 && echo '^^^ react-native in backend = BUG'
+grep -rn 'from.*generated/prisma' src/modules/ 2>/dev/null | head -3 && echo '^^^ prisma import (check if client exists)'
+echo '=== MISSING MODULE IMPORTS ==='
+grep -rn \"Cannot find module\" /app/output/*/generation.log 2>/dev/null | tail -5
+"
+```
+
+## Step 4: Code Substance Check (sample 5 files)
+
+For each: read first 20 lines, check if it has real logic (not just boilerplate)
+
+```bash
+docker exec coding-engine-api bash -c "
+cd /app/output/*/
+echo '=== SAMPLE: auth.controller.ts ==='
+head -25 src/modules/auth/auth.controller.ts 2>/dev/null || echo 'NOT FOUND'
+echo '=== SAMPLE: messages service ==='
+head -25 src/modules/messages/messages.service.ts 2>/dev/null || echo 'NOT FOUND'
+echo '=== SAMPLE: contacts controller ==='
+head -25 src/modules/contacts/contacts.controller.ts 2>/dev/null || echo 'NOT FOUND'
+echo '=== SAMPLE: chats service ==='
+head -25 src/modules/chats/chats.service.ts 2>/dev/null || echo 'NOT FOUND'
+echo '=== SAMPLE: users controller ==='
+head -25 src/modules/users/users.controller.ts 2>/dev/null || echo 'NOT FOUND'
+"
+```
+
+## Step 5: Build Test
+
+```bash
+docker exec coding-engine-api bash -c "
+cd /app/output/*/
+npx tsc --noEmit 2>&1 | tail -20
+echo '---EXIT CODE: '\$?
+"
+```
+
+## Step 6: Rate and Report
+
+Present findings as:
+
+```
+# 🔍 Code Quality Audit
+
+## File Inventory
+| Type | Count | With Real Logic | Empty/Stub |
+
+## Quality Rating: X/10
+- Real controllers with routes: X
+- Services with business logic: X
+- Empty/placeholder files: X
+- Broken imports: X
+- Build errors: X
+
+## Issues Found
+1. [CRITICAL] ...
+2. [WARNING] ...
+
+## Files That Need Regeneration
+- path/to/file.ts — reason
+```

.claude/skills/ce-health-check/SKILL.md

@@ -0,0 +1,94 @@
+---
+name: ce-health-check
+description: Quick health check of all Coding Engine infrastructure. Checks containers, DB connectivity, API endpoints, Discord bot, sandbox preview, LLM API keys, and generation process. Takes 15 seconds.
+trigger: When user asks "is everything running", "health check", "check infrastructure", "what's broken", "container status"
+---
+
+# ce-health-check — Infrastructure Health Check
+
+Run ALL checks in parallel for speed (target: 15 seconds).
+
+## Checks
+
+### 1. Containers
+```bash
+docker ps --format "{{.Names}} {{.Status}}" | grep coding-engine | sort
+```
+Expected: api (healthy), postgres (healthy), redis (healthy), sandbox (healthy), automation-ui (healthy), openclaw (healthy)
+
+### 2. API Health
+```bash
+curl -s --max-time 3 http://localhost:8000/health
+```
+Expected: `{"status":"healthy"}`
+
+### 3. DB Connectivity
+```bash
+docker exec coding-engine-postgres psql -U postgres -d coding_engine -c "SELECT COUNT(*) FROM tasks;" 2>&1
+```
+
+### 4. Redis
+```bash
+docker exec coding-engine-redis redis-cli ping
+```
+
+### 5. Sandbox Preview
+```bash
+curl -s --max-time 3 http://localhost:3100 2>/dev/null | head -1
+curl -s --max-time 3 http://localhost:6090 2>/dev/null | head -1
+```
+
+### 6. Discord Bot
+```bash
+docker logs coding-engine-automation-ui --tail 3 2>&1
+```
+
+### 7. LLM API Keys
+```bash
+docker exec coding-engine-api bash -c "
+echo 'OPENAI_API_KEY:' && [ -n \"\$OPENAI_API_KEY\" ] && echo 'SET' || echo 'MISSING'
+echo 'OPENROUTER_API_KEY:' && [ -n \"\$OPENROUTER_API_KEY\" ] && echo 'SET' || echo 'MISSING'
+echo 'ANTHROPIC_API_KEY:' && [ -n \"\$ANTHROPIC_API_KEY\" ] && echo 'SET' || echo 'MISSING'
+echo 'LLM_BACKEND:' \$LLM_BACKEND
+echo 'GITHUB_TOKEN:' && [ -n \"\$GITHUB_TOKEN\" ] && echo 'SET' || echo 'MISSING'
+echo 'DISCORD_BOT_TOKEN:' && [ -n \"\$DISCORD_BOT_TOKEN\" ] && echo 'SET' || echo 'MISSING'
+"
+```
+
+### 8. Generation Process
+```bash
+docker top coding-engine-api 2>/dev/null | grep run_generation | head -1 || echo "NO GENERATION RUNNING"
+```
+
+### 9. Disk Space
+```bash
+docker exec coding-engine-api bash -c "df -h /app/output | tail -1"
+```
+
+### 10. OpenRouter Credits (if used)
+```bash
+curl -s https://openrouter.ai/api/v1/auth/key \
+  -H "Authorization: Bearer $(docker exec coding-engine-api bash -c 'echo $OPENROUTER_API_KEY')" 2>/dev/null | \
+  python3 -c "import sys,json; d=json.load(sys.stdin).get('data',{}); print('Usage: \$%.2f' % d.get('usage',0))" 2>/dev/null || echo "Cannot check"
+```
+
+## Report Format
+
+```
+# 🏥 Health Check
+
+| Service | Status | Details |
+|---------|--------|---------|
+| API | ✅/❌ | |
+| Postgres | ✅/❌ | X tasks in DB |
+| Redis | ✅/❌ | |
+| Sandbox | ✅/❌ | Preview on :3100 |
+| VNC | ✅/❌ | Emulator on :6090 |
+| Discord Bot | ✅/❌ | |
+| Generation | 🔄/⏹️ | Round X/10 |
+| OpenAI Key | ✅/❌ | |
+| OpenRouter | ✅/❌ | $X used |
+| Disk | ✅/⚠️ | X% used |
+
+Issues: X found
+```

.claude/skills/ce-pipeline-doctor/SKILL.md

@@ -0,0 +1,167 @@
+---
+name: ce-pipeline-doctor
+description: Diagnoses pipeline problems and outputs actionable fix commands. Combines insights from status-report, code-quality, task-audit, and health-check into concrete recommendations with copy-paste-ready commands. Answers "what should I do next?" and "why is generation stuck?"
+trigger: When user asks "what's wrong", "fix the pipeline", "what should I do next", "why is it stuck", "diagnose", "doctor", "help me fix this"
+---
+
+# ce-pipeline-doctor — Pipeline Diagnosis & Action Plan
+
+Run a full diagnosis, then output a prioritized action plan with ready-to-execute commands.
+
+## Phase 1: Quick Triage (10 seconds)
+
+Run these checks to classify the problem:
+
+```bash
+# ONE-SHOT TRIAGE
+echo "=== CONTAINERS ===" && docker ps -a --format "{{.Names}} {{.Status}}" | grep coding-engine | sort
+echo "=== API HEALTH ===" && curl -s --max-time 3 http://localhost:8000/health 2>&1 || echo "API DOWN"
+echo "=== GENERATION ===" && docker top coding-engine-api 2>/dev/null | grep run_generation | head -1 || echo "NOT RUNNING"
+echo "=== DB TASKS ===" && docker exec coding-engine-postgres psql -U postgres -d coding_engine -t -c "SELECT status, COUNT(*) FROM tasks WHERE job_id=(SELECT MAX(id) FROM jobs) GROUP BY status ORDER BY status;" 2>&1
+echo "=== LAST ERROR ===" && docker exec coding-engine-api bash -c "grep 'ERROR\|OOM\|Kill\|Exit\|FAILED' /app/output/*/generation.log 2>/dev/null | grep -v 'subscriber\|entity_req\|_initialized\|enrichment' | tail -3" 2>&1
+echo "=== STALE RUNNING ===" && docker exec coding-engine-postgres psql -U postgres -d coding_engine -t -c "SELECT COUNT(*) FROM tasks WHERE job_id=(SELECT MAX(id) FROM jobs) AND status='RUNNING';" 2>&1
+echo "=== MEMORY ===" && docker stats coding-engine-api --no-stream --format "{{.MemUsage}}" 2>/dev/null || echo "Container not running"
+```
+
+## Phase 2: Classify Problem
+
+Based on triage results, classify into one of these categories:
+
+### Category A: Container Down
+- Symptom: API/Bot/Sandbox container Exited
+- Check exit code: 137=OOM, 139=SIGSEGV, 255=restart-failure
+
+### Category B: Generation Stuck/Dead
+- Symptom: `run_generation.py` not in process list but RUNNING tasks in DB
+- Stale RUNNING tasks = generation crashed mid-execution
+
+### Category C: Cascade Failure
+- Symptom: Many CANCELLED/SKIPPED tasks, few FAILED
+- Root cause: One upstream task failed → all dependents cancelled
+
+### Category D: LLM Issues
+- Symptom: 402 Payment Required, 429 Rate Limit, timeouts
+- Check which LLM backend and if credits are available
+
+### Category E: Task Explosion
+- Symptom: 2000+ PENDING tasks, generation runs but completes quickly with few results
+- Root cause: Tasks too granular, LLM calls wasted on tiny files
+
+### Category F: Code Quality Issues
+- Symptom: Build fails, wrong imports, empty generated files
+- Check: react-native in backend, missing prisma client, etc.
+
+## Phase 3: Generate Action Plan
+
+For each problem found, output a **numbered action** with the **exact command** to fix it.
+
+### Template Actions:
+
+#### Fix A: Container Down (OOM)
+```
+ACTION 1: Restart container with memory limit
+  docker compose up -d api
+
+ACTION 2: Set memory limit in docker-compose.yml
+  deploy:
+    resources:
+      limits:
+        memory: 4G
+
+ACTION 3: Reduce parallelism to lower memory usage
+  # In engine_settings.yml: generation.parallelism: 2 (was 3)
+```
+
+#### Fix B: Generation Stuck
+```
+ACTION 1: Reset stale RUNNING tasks
+  docker exec coding-engine-postgres psql -U postgres -d coding_engine -c "UPDATE tasks SET status='PENDING' WHERE job_id=(SELECT MAX(id) FROM jobs) AND status='RUNNING';"
+
+ACTION 2: Remove stale lock file
+  docker exec coding-engine-api bash -c "rm -f /app/output/*/.generation_running"
+
+ACTION 3: Restart generation
+  docker exec -d coding-engine-api bash -c "cd /app && python run_generation.py --project-path /app/Data/all_services/whatsapp-messaging-service_20260211_025459 --output-dir /app/output/whatsapp-messaging-service_20260211_025459 --project-id whatsapp-messaging-service --db-schema whatsapp_app --parallelism 2 --max-rounds 10 >> /app/output/whatsapp-messaging-service_20260211_025459/generation.log 2>&1"
+```
+
+#### Fix C: Cascade Failure
+```
+ACTION 1: Find cascade roots
+  docker exec coding-engine-postgres psql -U postgres -d coding_engine -c "SELECT task_id, status, status_message FROM tasks WHERE job_id=(SELECT MAX(id) FROM jobs) AND status='FAILED' LIMIT 10;"
+
+ACTION 2: Fix root failures (usually schema/migration)
+  # If migrations: mark as COMPLETED (schema already synced)
+  docker exec coding-engine-postgres psql -U postgres -d coding_engine -c "UPDATE tasks SET status='COMPLETED', status_message='Schema already synced' WHERE job_id=(SELECT MAX(id) FROM jobs) AND task_id LIKE '%migration%' AND status='FAILED';"
+
+ACTION 3: Reset cancelled tasks and regenerate
+  docker exec coding-engine-postgres psql -U postgres -d coding_engine -c "UPDATE tasks SET status='PENDING' WHERE job_id=(SELECT MAX(id) FROM jobs) AND status IN ('CANCELLED','SKIPPED');"
+```
+
+#### Fix D: LLM Issues
+```
+ACTION 1: Check API key balance
+  curl -s https://openrouter.ai/api/v1/auth/key -H "Authorization: Bearer $OPENROUTER_API_KEY" | python3 -c "import sys,json; print(json.load(sys.stdin)['data']['usage'])"
+
+ACTION 2: Switch to OpenAI if OpenRouter empty
+  # In docker-compose.yml: LLM_BACKEND=openai
+  docker compose up -d api
+
+ACTION 3: Use free models as fallback
+  # In engine_settings.yml: models.fixing.model: qwen/qwen3-coder:free
+```
+
+#### Fix E: Task Explosion
+```
+ACTION 1: Count the damage
+  docker exec coding-engine-postgres psql -U postgres -d coding_engine -t -c "SELECT 'API tasks: ' || COUNT(*) FROM tasks WHERE job_id=(SELECT MAX(id) FROM jobs) AND task_id LIKE '%API%';"
+
+ACTION 2: Mark redundant sub-tasks as COMPLETED
+  # Keep only -controller tasks, mark -guard, -validation as done
+  docker exec coding-engine-postgres psql -U postgres -d coding_engine -c "UPDATE tasks SET status='COMPLETED', status_message='Consolidated: generated with controller' WHERE job_id=(SELECT MAX(id) FROM jobs) AND (task_id LIKE '%-guard' OR task_id LIKE '%-validation') AND status='PENDING';"
+
+ACTION 3: For future projects: modify epic_task_generator.py
+  # Change _generate_api_tasks() to create 1 task per endpoint group instead of 5
+```
+
+#### Fix F: Code Quality
+```
+ACTION 1: Remove react-native imports from backend
+  docker exec coding-engine-api bash -c "cd /app/output/*/ && rm -f src/api/loginpageAPI.ts src/api/registerpageAPI.ts src/api/twofactorauthAPI.ts"
+
+ACTION 2: Generate prisma client
+  docker exec coding-engine-api bash -c "cd /app/output/*/ && DATABASE_URL='postgresql://postgres:postgres@postgres:5432/whatsapp_app?schema=public' npx prisma generate"
+
+ACTION 3: Run build check
+  docker exec coding-engine-api bash -c "cd /app/output/*/ && npx tsc --noEmit 2>&1 | head -20"
+```
+
+## Phase 4: Priority Matrix
+
+After generating actions, present them in priority order:
+
+```
+# 🏥 Pipeline Doctor — Diagnosis
+
+## Problem: [Category Name]
+[One sentence description]
+
+## Action Plan (in order):
+
+| # | Action | Impact | Command Ready? |
+|---|--------|--------|----------------|
+| 1 | [Most urgent] | HIGH | ✅ |
+| 2 | [Next] | HIGH | ✅ |
+| 3 | [Nice to have] | MEDIUM | ✅ |
+
+## Commands to Execute:
+[Copy-paste ready block of all commands in sequence]
+
+## Expected Outcome:
+- After Action 1: [what changes]
+- After Action 2: [what changes]
+- After all actions: [final state]
+```
+
+## Phase 5: Verify
+
+After actions are executed, re-run Phase 1 triage to confirm fixes worked.