-
Notifications
You must be signed in to change notification settings - Fork 3
Home
Multi-AI development pipelines with built-in security enforcement.
VCP is a protocol with three Claude Code plugins that orchestrate multiple AI models through structured development stages and enforce 41 security and quality standards with real-time blocking.
| Plugin | Purpose | Components |
|---|---|---|
| Dev Buddy | Multi-AI stage orchestration with Ralph loop architecture | 10 skills, 6 stage definitions, 6 role prompts |
| VCP | Standards enforcement — security, architecture, quality scanning | 10 skills, 1 agent, 4 hooks |
| MCP Doc | Documentation manifest generator for git-doc-mcp | 5 skills |
# 1. Add the VCP marketplace
/plugin marketplace add Z-M-Huang/vcp
# 2. Install plugins
/plugin install vcp@vcp
/plugin install vcp@dev-buddy
/plugin install vcp@mcp-doc
# 3. Initialize your project (VCP)
/vcp-init
# 4. Run your first audit
/vcp-audit src/
# 5. Build a feature with the Ralph loop
/dev-buddy-ralph Add user authenticationSee Getting Started for the full setup guide.
Dev Buddy uses the Ralph loop architecture — 6 stages with two nested loops:
Stage Definition + Role Prompt + AI Preset + Model = Executor
Executors are assigned to Stages (discovery, requirements, decomposition, build, code review, UAT)
Stage Skills dispatch executors with structural enforcement
| Skill | Command | Purpose |
|---|---|---|
| Ralph | /dev-buddy-ralph |
Full feature development — chains all 6 stages with loop logic |
| Discover | /dev-buddy-discover |
Explore codebase and running app |
| Requirements | /dev-buddy-requirements |
Define ACs + design UAT scenarios |
| Decompose | /dev-buddy-decompose |
Break feature into small testable units |
| Build | /dev-buddy-build |
Implement units with fresh context and backpressure |
| Code Review | /dev-buddy-code-review |
Multi-AI semantic drift detection with AC tracing |
| UAT | /dev-buddy-uat |
Execute UAT tests and full backpressure |
| Once | /dev-buddy-once |
Single task with any AI provider |
| Chatroom | /dev-buddy-chatroom |
Multi-AI competitive debate |
| Config | /dev-buddy-config |
Web configuration portal |
Anti-drift enforcement: Unit plan prompt, mechanical backpressure, orchestrator verify, multi-AI code review, UAT, user checkpoint, TaskManagement, plan files on disk.
See Dev Buddy Quick Start for details.
VCP enforces standards through three complementary layers:
| Layer | Mechanism | When |
|---|---|---|
| Proactive Context |
SessionStart hook + /vcp-context skill |
At session start and on demand |
| On-Demand Scanning | 10 skills | When you run them |
| Real-Time Blocking |
security-gate.ts hook |
Every Write/Edit/Bash call |
See Three‐Layer Enforcement Model for details.
- 10 skills — ralph, discover, requirements, decompose, build, code-review, uat, once, chatroom, config
- 6 stage definitions — discovery, ralph-requirements, decomposition, ralph-build, ralph-code-review, ralph-uat
- 6 role prompts — discoverer, ralph-requirements-analyst, decomposer, unit-builder, ralph-code-reviewer, uat-evaluator
- Ralph loop architecture — 2 nested loops (build inner, UAT outer) with unit-based decomposition
- Multi-provider support — Anthropic subscription, API (Anthropic/OpenAI compatible), CLI tools (Codex)
-
Structural enforcement —
--allowed-toolsfor API executors, stage gate hook, TaskManagement
- 41 standards across 12 scopes
- 10 skills — audit, dependency-check, pre-commit-review, review-tests, coverage-gaps, test-plan, root-cause-check, config, context, init
-
1 agent —
migration-planner(phased remediation planning) - 4 hooks — security-context, security-gate (21 regex patterns, 9 CWEs), test-quality-warning, stop-reminder
- First-Time Setup Guide — Step-by-step from install to first run
- How Configuration Works — What each setting does, decision guidance
- Configuration Recipes — Ready-to-use configs for solo dev, team, compliance
- Web Portal Guide — Tab-by-tab walk-through of the config portal
- Daily VCP Workflow — When to use which skill, recommended daily flow
- Troubleshooting — Common errors, debug mode, recovery
- Dev Buddy Quick Start — Installation, first Ralph loop run
- Dev Buddy Configuration — Stages, executors, settings (v5 config)
- Stage Skills Guide — Ralph loop architecture and how each stage works
- AI Provider Presets — Subscription, API, and CLI presets
- System Prompts Reference — Stage definitions + role prompts
- Chatroom — Multi-model deliberation
-
Configuration —
.vcp/config.jsonschema, scopes, compliance - Skills Reference — All 10 skills with usage and output formats
- Three‐Layer Enforcement Model — How proactive context, scanning, and blocking work
- Hooks Reference — VCP hooks (4) and Dev Buddy hooks (3)
- Security Gate Patterns — All 21 regex patterns across 9 CWEs
- Shared Modules — Context generation, config resolution
- Standards Overview — All 41 standards organized by scope
- Standards Format Specification — How to write VCP-compliant standards
- Roadmap — Current status and planned features
- FAQ — Common questions
- Contributing — How to report problems and submit PRs
- Known Limitations — What VCP can and cannot do today
VCP Wiki
Guides
- First-Time Setup Guide
- How Configuration Works
- Configuration Recipes
- Web Portal Guide
- Daily VCP Workflow
- Troubleshooting
VCP Plugin
- Configuration
- Skills Reference
- Three‐Layer Enforcement Model
- Hooks Reference
- Security Gate Patterns
- Shared Modules
Dev Buddy Plugin
- Dev Buddy Quick Start
- Dev Buddy Configuration
- Stage Skills Guide
- AI Provider Presets
- System Prompts Reference
- Chatroom
MCP Doc Plugin
Standards
Project
VCP Wiki (中文)
指南
VCP 插件
Dev Buddy 插件
MCP Doc 插件
标准
项目