Agent Skills › ruvnet/RuVector › Custom Workers

Custom Workers

GitHub

构建可组合的后台分析工作流，支持代码扫描、安全检测、模式学习及API文档生成。提供多种预设和24种阶段执行器，用于自动化静态分析、漏洞发现及向量嵌入处理。

.claude/skills/custom-workers/SKILL.md ruvnet/RuVector

Trigger Scenarios

需要自动化代码质量或安全扫描生成API文档进行代码模式学习与记忆

Install

npx skills add ruvnet/RuVector --skill Custom Workers -g -y

More Options

Non-standard path

npx skills add https://github.com/ruvnet/RuVector/tree/main/.claude/skills/custom-workers -g -y

Use without installing

npx skills use ruvnet/RuVector@Custom Workers

指定 Agent (Claude Code)

npx skills add ruvnet/RuVector --skill Custom Workers -a claude-code -g -y

安装 repo 全部 skill

npx skills add ruvnet/RuVector --all -g -y

预览 repo 内 skill

npx skills add ruvnet/RuVector --list

SKILL.md

Frontmatter

{
    "name": "Custom Workers",
    "hooks": {
        "pre": "echo \"🔧 Custom Workers activated\"\n.claude\/agentic-flow-fast.sh workers stats 2>\/dev\/null || true\n",
        "post": "echo \"✅ Custom Workers complete\"\n.claude\/agentic-flow-fast.sh workers results --json 2>\/dev\/null | head -20 || true"
    },
    "description": "Create and run custom background analysis workers with composable phases. Use when you need automated code analysis, security scanning, pattern learning, or API documentation generation."
}

Custom Workers

Build composable background analysis workers with 24 phase executors and 6 presets.

Quick Start

# List available presets
npx ruvector workers presets

# List available phase executors
npx ruvector workers phases

# Create a custom worker from preset
npx ruvector workers create my-scanner --preset security-scan

# Run the worker
npx ruvector workers run my-scanner --path ./src

Available Presets

Preset	Description	Phases
`quick-scan`	Fast file discovery and stats	file-discovery → summarization
`deep-analysis`	Comprehensive code analysis	file-discovery → static-analysis → complexity-analysis → import-analysis → pattern-extraction → graph-build → vectorization → summarization
`security-scan`	Security-focused analysis	file-discovery → security-analysis → secret-detection → dependency-discovery → report-generation
`learning`	Pattern learning and memory	file-discovery → pattern-extraction → embedding-generation → pattern-storage → sona-training
`api-docs`	API endpoint documentation	file-discovery → api-discovery → type-analysis → report-generation
`test-analysis`	Test coverage analysis	file-discovery → static-analysis → pattern-extraction → summarization

Phase Executors (24 total)

Discovery Phases

file-discovery - Find files matching patterns
pattern-discovery - Discover code patterns
dependency-discovery - Map dependencies
api-discovery - Find API endpoints

Analysis Phases

static-analysis - AST-based code analysis
complexity-analysis - Cyclomatic complexity
security-analysis - Security vulnerability scan
performance-analysis - Performance bottlenecks
import-analysis - Import/export mapping
type-analysis - TypeScript type extraction

Pattern Phases

pattern-extraction - Extract code patterns
todo-extraction - Find TODOs/FIXMEs
secret-detection - Detect hardcoded secrets
code-smell-detection - Identify code smells

Build Phases

graph-build - Build code graph
call-graph - Function call graph
dependency-graph - Dependency graph

Learning Phases

vectorization - Convert to vectors
embedding-generation - ONNX embeddings (384d)
pattern-storage - Store in VectorDB
sona-training - SONA reinforcement learning

Output Phases

summarization - Generate summary
report-generation - Create report
indexing - Index for search

YAML Configuration

Create workers.yaml in your project:

version: '1.0'
workers:
  - name: auth-scanner
    triggers: [auth-scan, scan-auth]
    phases:
      - type: file-discovery
        config:
          patterns: ["**/auth/**", "**/login/**"]
      - type: security-analysis
      - type: secret-detection
      - type: report-generation
    capabilities:
      onnxEmbeddings: true
      vectorDb: true

  - name: api-documenter
    triggers: [api-docs, document-api]
    phases:
      - type: file-discovery
        config:
          patterns: ["**/routes/**", "**/api/**"]
      - type: api-discovery
      - type: type-analysis
      - type: report-generation

Load configuration:

npx ruvector workers load-config --file workers.yaml

CLI Commands

# Core commands
npx ruvector workers presets       # List presets
npx ruvector workers phases        # List phases
npx ruvector workers create <name> --preset <preset>
npx ruvector workers run <name> --path <path>

# Configuration
npx ruvector workers init-config   # Generate workers.yaml
npx ruvector workers load-config   # Load from workers.yaml
npx ruvector workers custom        # List registered workers

# Monitoring
npx ruvector workers status        # Worker status
npx ruvector workers results       # Analysis results
npx ruvector workers stats         # Statistics

MCP Tools

Available via ruvector MCP server:

Tool	Description
`workers_presets`	List available presets
`workers_phases`	List phase executors
`workers_create`	Create custom worker
`workers_run`	Run worker on path
`workers_custom`	List custom workers
`workers_init_config`	Generate config
`workers_load_config`	Load config

Capabilities

Workers can use these capabilities:

ONNX Embeddings: Real transformer embeddings (all-MiniLM-L6-v2, 384d, SIMD)
VectorDB: Store and search embeddings with HNSW indexing
SONA Learning: Self-Organizing Neural Architecture for pattern learning
ReasoningBank: Trajectory tracking and meta-learning

Integration with Hooks

Workers auto-dispatch on UserPromptSubmit via trigger keywords:

Type "audit this code" → triggers audit worker
Type "security scan" → triggers security worker
Type "learn patterns" → triggers learning worker

Example: Security Scanner

# Create from security-scan preset
npx ruvector workers create security-checker --preset security-scan --triggers "security,vuln,cve"

# Run on source
npx ruvector workers run security-checker --path ./src

# View results
npx ruvector workers results

Example: Custom Learning Worker

# workers.yaml
workers:
  - name: code-learner
    triggers: [learn, pattern-learn]
    phases:
      - type: file-discovery
        config:
          patterns: ["**/*.ts", "**/*.js"]
          exclude: ["node_modules/**"]
      - type: static-analysis
      - type: pattern-extraction
      - type: embedding-generation
      - type: pattern-storage
      - type: sona-training
    capabilities:
      onnxEmbeddings: true
      vectorDb: true
      sonaLearning: true