Coordinación, formateo y aprendizaje automáticos de las operaciones de Claude Code mediante hooks inteligentes integrados con MCP.
en todo el repo
0–100, la ruta de este skill
último commit aquí
últimos 90 días
69 tok en reposo
31 KB
Funciona con cualquier agente que lea SKILL.md
npx -y skills add ruvnet/ruflo --skill hooks-automation --agent claude-codeSe instala solo en este repositorio.
Este skill reads environment config.
Di cualquiera de estas frases y el agente debería cargar este skill.
Intelligent automation system that coordinates, validates, and learns from Claude Code operations through hooks integrated with MCP tools and neural pattern training.
This skill provides a comprehensive hook system that automatically manages development operations, coordinates swarm agents, maintains session state, and continuously learns from coding patterns. It enables automated agent assignment, code formatting, performance tracking, and cross-session memory persistence.
Key Capabilities:
Required:
npm install -g claude-flow@alpha).claude$settings.json with hook configurationsOptional:
# Initialize with default hooks configuration
npx claude-flow init --hooks
This creates:
.claude$settings.json with pre-configured hooks.claude$commands$hooks/# Pre-task hook (auto-spawns agents)
npx claude-flow hook pre-task --description "Implement authentication"
# Post-edit hook (auto-formats and stores in memory)
npx claude-flow hook post-edit --file "src$auth.js" --memory-key "auth$login"
# Session end hook (saves state and metrics)
npx claude-flow hook session-end --session-id "dev-session" --export-metrics
Hooks that execute BEFORE operations to prepare and validate:
pre-edit - Validate and assign agents before file modifications
npx claude-flow hook pre-edit [options]
Options:
--file, -f <path> File path to be edited
--auto-assign-agent Automatically assign best agent (default: true)
--validate-syntax Pre-validate syntax before edit
--check-conflicts Check for merge conflicts
--backup-file Create backup before editing
Examples:
npx claude-flow hook pre-edit --file "src$auth$login.js"
npx claude-flow hook pre-edit -f "config$db.js" --validate-syntax
npx claude-flow hook pre-edit -f "production.env" --backup-file --check-conflicts
Features:
pre-bash - Check command safety and resource requirements
npx claude-flow hook pre-bash --command <cmd>
Options:
--command, -c <cmd> Command to validate
--check-safety Verify command safety (default: true)
--estimate-resources Estimate resource usage
--require-confirmation Request user confirmation for risky commands
Examples:
npx claude-flow hook pre-bash -c "rm -rf /tmp/cache"
npx claude-flow hook pre-bash --command "docker build ." --estimate-resources
Features:
pre-task - Auto-spawn agents and prepare for complex tasks
npx claude-flow hook pre-task [options]
Options:
--description, -d <text> Task description for context
--auto-spawn-agents Automatically spawn required agents (default: true)
--load-memory Load relevant memory from previous sessions
--optimize-topology Select optimal swarm topology
--estimate-complexity Analyze task complexity
Examples:
npx claude-flow hook pre-task --description "Implement user authentication"
npx claude-flow hook pre-task -d "Continue API dev" --load-memory
npx claude-flow hook pre-task -d "Refactor codebase" --optimize-topology
Features:
pre-search - Prepare and optimize search operations
npx claude-flow hook pre-search --query <query>
Options:
--query, -q <text> Search query
--check-cache Check cache first (default: true)
--optimize-query Optimize search pattern
Examples:
npx claude-flow hook pre-search -q "authentication middleware"
Features:
Hooks that execute AFTER operations to process and learn:
post-edit - Auto-format, validate, and update memory
npx claude-flow hook post-edit [options]
Options:
--file, -f <path> File path that was edited
--auto-format Automatically format code (default: true)
--memory-key, -m <key> Store edit context in memory
--train-patterns Train neural patterns from edit
--validate-output Validate edited file
Examples:
npx claude-flow hook post-edit --file "src$components/Button.jsx"
npx claude-flow hook post-edit -f "api$auth.js" --memory-key "auth$login"
npx claude-flow hook post-edit -f "utils$helpers.ts" --train-patterns
Features:
post-bash - Log execution and update metrics
npx claude-flow hook post-bash --command <cmd>
Options:
--command, -c <cmd> Command that was executed
--log-output Log command output (default: true)
--update-metrics Update performance metrics
--store-result Store result in memory
Examples:
npx claude-flow hook post-bash -c "npm test" --update-metrics
Features:
post-task - Performance analysis and decision storage
npx claude-flow hook post-task [options]
Options:
--task-id, -t <id> Task identifier for tracking
--analyze-performance Generate performance metrics (default: true)
--store-decisions Save task decisions to memory
--export-learnings Export neural pattern learnings
--generate-report Create task completion report
Examples:
npx claude-flow hook post-task --task-id "auth-implementation"
npx claude-flow hook post-task -t "api-refactor" --analyze-performance
npx claude-flow hook post-task -t "bug-fix-123" --store-decisions
Features:
post-search - Cache results and improve patterns
npx claude-flow hook post-search --query <query> --results <path>
Options:
--query, -q <text> Original search query
--results, -r <path> Results file path
--cache-results Cache for future use (default: true)
--train-patterns Improve search patterns
Examples:
npx claude-flow hook post-search -q "auth" -r "results.json" --train-patterns
Features:
Hooks that coordinate with MCP swarm tools:
mcp-initialized - Persist swarm configuration
npx claude-flow hook mcp-initialized --swarm-id <id>
Features:
- Save swarm topology and configuration
- Store agent roster in memory
- Initialize coordination namespace
agent-spawned - Update agent roster and memory
npx claude-flow hook agent-spawned --agent-id <id> --type <type>
Features:
- Register agent in coordination memory
- Update agent roster
- Initialize agent-specific memory namespace
task-orchestrated - Monitor task progress
npx claude-flow hook task-orchestrated --task-id <id>
Features:
- Track task progress through memory
- Monitor agent assignments
- Update coordination state
neural-trained - Save pattern improvements
npx claude-flow hook neural-trained --pattern <name>
Features:
- Export trained neural patterns
- Update coordination models
- Share learning across agents
memory-write - Triggered when agents write to coordination memory
Features:
- Validate memory key format
- Update cross-agent indexes
- Trigger dependent hooks
- Notify subscribed agents
memory-read - Triggered when agents read from coordination memory
Features:
- Log access patterns
- Update popularity metrics
- Preload related data
- Track usage statistics
memory-sync - Synchronize memory across swarm agents
npx claude-flow hook memory-sync --namespace <ns>
Features:
- Sync memory state across agents
- Resolve conflicts
- Propagate updates
- Maintain consistency
session-start - Initialize new session
npx claude-flow hook session-start --session-id <id>
Options:
--session-id, -s <id> Session identifier
--load-context Load context from previous session
--init-agents Initialize required agents
Features:
- Create session directory
- Initialize metrics tracking
- Load previous context
- Set up coordination namespace
session-restore - Load previous session state
npx claude-flow hook session-restore --session-id <id>
Options:
--session-id, -s <id> Session to restore
--restore-memory Restore memory state (default: true)
--restore-agents Restore agent configurations
Examples:
npx claude-flow hook session-restore --session-id "swarm-20241019"
npx claude-flow hook session-restore -s "feature-auth" --restore-memory
Features:
session-end - Cleanup and persist session state
npx claude-flow hook session-end [options]
Options:
--session-id, -s <id> Session identifier to end
--save-state Save current session state (default: true)
--export-metrics Export session metrics
--generate-summary Create session summary
--cleanup-temp Remove temporary files
Examples:
npx claude-flow hook session-end --session-id "dev-session-2024"
npx claude-flow hook session-end -s "feature-auth" --export-metrics --generate-summary
npx claude-flow hook session-end -s "quick-fix" --cleanup-temp
Features:
notify - Custom notifications with swarm status
npx claude-flow hook notify --message <msg>
Options:
--message, -m <text> Notification message
--level <level> Notification level (info|warning|error)
--swarm-status Include swarm status (default: true)
--broadcast Send to all agents
Examples:
npx claude-flow hook notify -m "Task completed" --level info
npx claude-flow hook notify -m "Critical error" --level error --broadcast
Features:
Edit .claude$settings.json to configure hooks:
{
"hooks": {
"PreToolUse": [
{
"matcher": "^(Write|Edit|MultiEdit)$",
"hooks": [{
"type": "command",
"command": "npx claude-flow hook pre-edit --file '${tool.params.file_path}' --memory-key 'swarm$editor$current'"
}]
},
{
"matcher": "^Bash$",
"hooks": [{
"type": "command",
"command": "npx claude-flow hook pre-bash --command '${tool.params.command}'"
}]
}
],
"PostToolUse": [
{
"matcher": "^(Write|Edit|MultiEdit)$",
"hooks": [{
"type": "command",
"command": "npx claude-flow hook post-edit --file '${tool.params.file_path}' --memory-key 'swarm$editor$complete' --auto-format --train-patterns"
}]
},
{
"matcher": "^Bash$",
"hooks": [{
"type": "command",
"command": "npx claude-flow hook post-bash --command '${tool.params.command}' --update-metrics"
}]
}
]
}
}
Complete hook configuration with all features:
{
"hooks": {
"enabled": true,
"debug": false,
"timeout": 5000,
"PreToolUse": [
{
"matcher": "^(Write|Edit|MultiEdit)$",
"hooks": [
{
"type": "command",
"command": "npx claude-flow hook pre-edit --file '${tool.params.file_path}' --auto-assign-agent --validate-syntax",
"timeout": 3000,
"continueOnError": true
}
]
},
{
"matcher": "^Task$",
"hooks": [
{
"type": "command",
"command": "npx claude-flow hook pre-task --description '${tool.params.task}' --auto-spawn-agents --load-memory",
"async": true
}
]
},
{
"matcher": "^Grep$",
"hooks": [
{
"type": "command",
"command": "npx claude-flow hook pre-search --query '${tool.params.pattern}' --check-cache"
}
]
}
],
"PostToolUse": [
{
"matcher": "^(Write|Edit|MultiEdit)$",
"hooks": [
{
"type": "command",
"command": "npx claude-flow hook post-edit --file '${tool.params.file_path}' --memory-key 'edits/${tool.params.file_path}' --auto-format --train-patterns",
"async": true
}
]
},
{
"matcher": "^Task$",
"hooks": [
{
"type": "command",
"command": "npx claude-flow hook post-task --task-id '${result.task_id}' --analyze-performance --store-decisions --export-learnings",
"async": true
}
]
},
{
"matcher": "^Grep$",
"hooks": [
{
"type": "command",
"command": "npx claude-flow hook post-search --query '${tool.params.pattern}' --cache-results --train-patterns"
}
]
}
],
"SessionStart": [
{
"hooks": [
{
"type": "command",
"command": "npx claude-flow hook session-start --session-id '${session.id}' --load-context"
}
]
}
],
"SessionEnd": [
{
"hooks": [
{
"type": "command",
"command": "npx claude-flow hook session-end --session-id '${session.id}' --export-metrics --generate-summary --cleanup-temp"
}
]
}
]
}
}
Add protection for sensitive files:
{
"hooks": {
"PreToolUse": [
{
"matcher": "^(Write|Edit|MultiEdit)$",
"hooks": [
{
"type": "command",
"command": "npx claude-flow hook check-protected --file '${tool.params.file_path}'"
}
]
}
]
}
}
Run tests after file modifications:
{
"hooks": {
"PostToolUse": [
{
"matcher": "^Write$",
"hooks": [
{
"type": "command",
"command": "test -f '${tool.params.file_path%.js}.test.js' && npm test '${tool.params.file_path%.js}.test.js'",
"continueOnError": true
}
]
}
]
}
}
Hooks automatically integrate with MCP tools for coordination:
// Hook command
npx claude-flow hook pre-task --description "Build REST API"
// Internally calls MCP tools:
mcp__claude-flow__agent_spawn {
type: "backend-dev",
capabilities: ["api", "database", "testing"]
}
mcp__claude-flow__memory_usage {
action: "store",
key: "swarm$task$api-build$context",
namespace: "coordination",
value: JSON.stringify({
description: "Build REST API",
agents: ["backend-dev"],
started: Date.now()
})
}
// Hook command
npx claude-flow hook post-edit --file "api$auth.js"
// Internally calls MCP tools:
mcp__claude-flow__memory_usage {
action: "store",
key: "swarm$edits$api$auth.js",
namespace: "coordination",
value: JSON.stringify({
file: "api$auth.js",
timestamp: Date.now(),
changes: { added: 45, removed: 12 },
formatted: true,
linted: true
})
}
mcp__claude-flow__neural_train {
pattern_type: "coordination",
training_data: { /* edit patterns */ }
}
// Hook command
npx claude-flow hook session-end --session-id "dev-2024"
// Internally calls MCP tools:
mcp__claude-flow__memory_persist {
sessionId: "dev-2024"
}
mcp__claude-flow__swarm_status {
swarmId: "current"
}
// Generates metrics and summary
All hooks follow a standardized memory coordination pattern:
Phase 1: STATUS - Hook starts
mcp__claude-flow__memory_usage {
action: "store",
key: "swarm$hooks$pre-edit$status",
namespace: "coordination",
value: JSON.stringify({
status: "running",
hook: "pre-edit",
file: "src$auth.js",
timestamp: Date.now()
})
}
Phase 2: PROGRESS - Hook processes
mcp__claude-flow__memory_usage {
action: "store",
key: "swarm$hooks$pre-edit$progress",
namespace: "coordination",
value: JSON.stringify({
progress: 50,
action: "validating syntax",
file: "src$auth.js"
})
}
Phase 3: COMPLETE - Hook finishes
mcp__claude-flow__memory_usage {
action: "store",
key: "swarm$hooks$pre-edit$complete",
namespace: "coordination",
value: JSON.stringify({
status: "complete",
result: "success",
agent_assigned: "backend-dev",
syntax_valid: true,
backup_created: true
})
}
Hooks return JSON responses to control operation flow:
{
"continue": true,
"reason": "All validations passed",
"metadata": {
"agent_assigned": "backend-dev",
"syntax_valid": true,
"file": "src$auth.js"
}
}
{
"continue": false,
"reason": "Protected file - manual review required",
"metadata": {
"file": ".env.production",
"protection_level": "high",
"requires": "manual_approval"
}
}
{
"continue": true,
"reason": "Syntax valid but complexity high",
"warnings": [
"Cyclomatic complexity: 15 (threshold: 10)",
"Consider refactoring for better maintainability"
],
"metadata": {
"complexity": 15,
"threshold": 10
}
}
Hooks can integrate with Git operations for quality control:
# Add to .git$hooks$pre-commit or use husky
#!$bin$bash
# Run quality checks before commit
# Get staged files
FILES=$(git diff --cached --name-only --diff-filter=ACM)
for FILE in $FILES; do
# Run pre-edit hook for validation
npx claude-flow hook pre-edit --file "$FILE" --validate-syntax
if [ $? -ne 0 ]; then
echo "Validation failed for $FILE"
exit 1
fi
# Run post-edit hook for formatting
npx claude-flow hook post-edit --file "$FILE" --auto-format
done
# Run tests
npm test
exit $?
# Add to .git$hooks$post-commit
#!$bin$bash
# Track commit metrics
COMMIT_HASH=$(git rev-parse HEAD)
COMMIT_MSG=$(git log -1 --pretty=%B)
npx claude-flow hook notify \
--message "Commit completed: $COMMIT_MSG" \
--level info \
--swarm-status
# Add to .git$hooks$pre-push
#!$bin$bash
# Quality gate before push
# Run full test suite
npm run test:all
# Run quality checks
npx claude-flow hook session-end \
--generate-report \
--export-metrics
# Verify quality thresholds
TRUTH_SCORE=$(npx claude-flow metrics score --format json | jq -r '.truth_score')
if (( $(echo "$TRUTH_SCORE < 0.95" | bc -l) )); then
echo "Truth score below threshold: $TRUTH_SCORE < 0.95"
exit 1
fi
exit 0
How agents use hooks for coordination:
# Agent 1: Backend Developer
# STEP 1: Pre-task preparation
npx claude-flow hook pre-task \
--description "Implement user authentication API" \
--auto-spawn-agents \
--load-memory
# STEP 2: Work begins - pre-edit validation
npx claude-flow hook pre-edit \
--file "api$auth.js" \
--auto-assign-agent \
--validate-syntax
# STEP 3: Edit file (via Claude Code Edit tool)
# ... code changes ...
# STEP 4: Post-edit processing
npx claude-flow hook post-edit \
--file "api$auth.js" \
--memory-key "swarm$backend$auth-api" \
--auto-format \
--train-patterns
# STEP 5: Notify coordination system
npx claude-flow hook notify \
--message "Auth API implementation complete" \
--swarm-status \
--broadcast
# STEP 6: Task completion
npx claude-flow hook post-task \
--task-id "auth-api" \
--analyze-performance \
--store-decisions \
--export-learnings
# Agent 2: Test Engineer (receives notification)
# STEP 1: Check memory for API details
npx claude-flow hook session-restore \
--session-id "swarm-current" \
--restore-memory
# Memory contains: swarm$backend$auth-api with implementation details
# STEP 2: Generate tests
npx claude-flow hook pre-task \
--description "Write tests for auth API" \
--load-memory
# STEP 3: Create test file
npx claude-flow hook post-edit \
--file "api$auth.test.js" \
--memory-key "swarm$testing$auth-api-tests" \
--train-patterns
# STEP 4: Share test results
npx claude-flow hook notify \
--message "Auth API tests complete - 100% coverage" \
--broadcast
Create custom hooks for specific workflows:
// .claude$hooks$custom-quality-check.js
module.exports = {
name: 'custom-quality-check',
type: 'pre',
matcher: /\.(ts|js)$/,
async execute(context) {
const { file, content } = context;
// Custom validation logic
const complexity = await analyzeComplexity(content);
const securityIssues = await scanSecurity(content);
// Store in memory
await storeInMemory({
key: `quality/${file}`,
value: { complexity, securityIssues }
});
// Return decision
if (complexity > 15 || securityIssues.length > 0) {
return {
continue: false,
reason: 'Quality checks failed',
warnings: [
`Complexity: ${complexity} (max: 15)`,
`Security issues: ${securityIssues.length}`
]
};
}
return {
continue: true,
reason: 'Quality checks passed',
metadata: { complexity, securityIssues: 0 }
};
}
};
{
"hooks": {
"PreToolUse": [
{
"matcher": "^(Write|Edit)$",
"hooks": [
{
"type": "script",
"script": ".claude$hooks$custom-quality-check.js"
}
]
}
]
}
}
# Session start - initialize coordination
npx claude-flow hook session-start --session-id "fullstack-feature"
# Pre-task planning
npx claude-flow hook pre-task \
--description "Build user profile feature - frontend + backend + tests" \
--auto-spawn-agents \
--optimize-topology
# Backend work
npx claude-flow hook pre-edit --file "api$profile.js"
# ... implement backend ...
npx claude-flow hook post-edit \
--file "api$profile.js" \
--memory-key "profile$backend" \
--train-patterns
# Frontend work (reads backend details from memory)
npx claude-flow hook pre-edit --file "components/Profile.jsx"
# ... implement frontend ...
npx claude-flow hook post-edit \
--file "components/Profile.jsx" \
--memory-key "profile$frontend" \
--train-patterns
# Testing (reads both backend and frontend from memory)
npx claude-flow hook pre-task \
--description "Test profile feature" \
--load-memory
# Session end - export everything
npx claude-flow hook session-end \
--session-id "fullstack-feature" \
--export-metrics \
--generate-summary
# Start debugging session
npx claude-flow hook session-start --session-id "debug-memory-leak"
# Pre-task: analyze issue
npx claude-flow hook pre-task \
--description "Debug memory leak in event handlers" \
--load-memory \
--estimate-complexity
# Search for event emitters
npx claude-flow hook pre-search --query "EventEmitter"
# ... search executes ...
npx claude-flow hook post-search \
--query "EventEmitter" \
--cache-results
# Fix the issue
npx claude-flow hook pre-edit \
--file "services$events.js" \
--backup-file
# ... fix code ...
npx claude-flow hook post-edit \
--file "services$events.js" \
--memory-key "debug$memory-leak-fix" \
--validate-output
# Verify fix
npx claude-flow hook post-task \
--task-id "memory-leak-fix" \
--analyze-performance \
--generate-report
# End session
npx claude-flow hook session-end \
--session-id "debug-memory-leak" \
--export-metrics
# Initialize swarm for refactoring
npx claude-flow hook pre-task \
--description "Refactor legacy codebase to modern patterns" \
--auto-spawn-agents \
--optimize-topology
# Agent 1: Code Analyzer
npx claude-flow hook pre-task --description "Analyze code complexity"
# ... analysis ...
npx claude-flow hook post-task \
--task-id "analysis" \
--store-decisions
# Agent 2: Refactoring (reads analysis from memory)
npx claude-flow hook session-restore \
--session-id "swarm-refactor" \
--restore-memory
for file in src/**/*.js; do
npx claude-flow hook pre-edit --file "$file" --backup-file
# ... refactor ...
npx claude-flow hook post-edit \
--file "$file" \
--memory-key "refactor/$file" \
--auto-format \
--train-patterns
done
# Agent 3: Testing (reads refactored code from memory)
npx claude-flow hook pre-task \
--description "Generate tests for refactored code" \
--load-memory
# Broadcast completion
npx claude-flow hook notify \
--message "Refactoring complete - all tests passing" \
--broadcast
Enable debug mode for troubleshooting:
# Enable debug output
export CLAUDE_FLOW_DEBUG=true
# Test specific hook with verbose output
npx claude-flow hook pre-edit --file "test.js" --debug
# Check hook execution logs
cat .claude-flow$logs$hooks-$(date +%Y-%m-%d).log
# Validate configuration
npx claude-flow hook validate-config
.claude$settings.json syntaxnpx claude-flow init --hooks - Initialize hooks systemnpx claude-flow hook --list - List available hooksnpx claude-flow hook --test <hook> - Test specific hooknpx claude-flow memory usage - Manage memorynpx claude-flow agent spawn - Spawn agentsnpx claude-flow swarm init - Initialize swarmThis skill works seamlessly with:
Reproducido de ruvnet/ruflo bajo licencia MIT. Leer esta página en markdown.
1 archivo en el paquete. Solo se lee SKILL.md al activarse — las referencias se cargan si el skill decide que las necesita.
Requiere Claude Flow CLI (npm install -g claude-flow@alpha) y hooks habilitados en .claude/settings.json.
Necesita en el PATH:gitjqnpmnpx
Variables de entorno:COMMIT_MSGFILEFILESTRUTH_SCORE
Este repo incluye 275 skills. Si instalas uno, normalmente ya tienes los demás.
Inspecciona y audita genomas GEPA: carga y valida un genoma, renderiza el system prompt que compila, o clasifica los modos de fallo de una transcripción de ejecución.
Completion en modo razonamiento contra deepseek-reasoner (R1) de DeepSeek. Devuelve el chain-of-thought por separado de la respuesta final. Lee DEEPSEEK_API_KEY y degrada si falta o la API no responde.
Completion de un solo turno contra el modelo deepseek-chat de DeepSeek vía /v1/chat/completions. Lee DEEPSEEK_API_KEY y degrada con status:degraded si falta o la API no responde. Para tareas sin razonamiento.
Construye o reconstruye el índice de ADRs y su grafo de dependencias ejecutando scripts/import.mjs, en vez de cientos de llamadas MCP.
Muestra el estado de la integración AGNTCY/SLIM/CASA: si los paquetes están instalados, qué transporte está activo y si el enforcement de CASA está habilitado.
Ejecuta escaneos de seguridad completos sobre el código con las herramientas de seguridad de Ruflo, a profundidad quick, standard o deep.
