VClaw Continuous Development & Automation Blueprint

This document outlines the strategy for the Continuous, Automated, and AI-native development of VClaw. As a "controlled product fork" of OpenClaw, VClaw requires a specialized workflow that keeps the core up-to-date while automating the evolution of the business layer in the direction of a growth + operations assistant.

1. Automated Infrastructure

1.1 Git Submodule Strategy

• Upstream Sync: Weekly automated pulls from openclaw/openclaw into /core/openclaw.
• Plugin Decoupling: Business logic (VietQR, Shipping, content assistance, follow-up, etc.) is developed as independent OpenClaw plugins to avoid merge conflicts with the core.
• VClaw UI: Versioned independently, consuming core APIs via standard WebSocket/REST contracts.

1.2 Development Environment Automation

• Bootstrap Script: A single scripts/bootstrap.mjs that:
  1. Installs all UI dependencies.
  2. Runs pnpm install and pnpm build in the core.
  3. Verifies the Node.js and Go environment.
• One-Click Startup: npm run dev:full will concurrently start the OpenClaw Gateway (core) and the Next.js Operations Console (UI).

2. Agentic Development Workflow (Coding via Messages)

VClaw is designed to be Agentic-first. This means the software itself helps you build it.

2.1 Workspace Awareness

OpenClaw is configured with a workspace.json that gives the internal AI agent full context of:

• The vclaw-ui component library (shadcn/ui).
• The core/openclaw plugin API.
• The docs/ directory as the "Source of Truth" for requirements.

2.2 Continuous Coding Loop

1. Instruction: You send a message via the VClaw Admin Chat (e.g., "VClaw, add a content-approval section for follow-up campaigns to the admin workspace").
2. Analysis: The VClaw Dev Agent analyzes the React component and the underlying JSON schema in the Core.
3. Execution: The Agent performs the multi-file edit (UI form, i18n payloads, automation policy, and backend schema).
4. Verification: Automated Vitest suites run to ensure no regressions.

3. CI/CD & Deployment Automation

3.1 Automated Quality Assurance

• Lint Gating: Pre-commit hooks run oxlint and eslint to maintain banking-grade code quality.
• Agentic Testing: If a test fails, a specialized "Fixer Agent" investigates the logs and proposes a patch.

3.2 Continuous Delivery

• Vercel Integration: Automatic previews of the vclaw-ui surface for every PR.
• Native Packaging: Github Actions automated nightly builds of the .exe and .dmg 1-click installers.

4. Roadmap to Full Automation

[!IMPORTANT]
All automated "Self-Improvement" actions, outbound drafts, follow-up automation, or guarded consultation flows must be gated by the Human-in-the-loop Task Inbox or an explicitly configured policy. AI never pushes code to production or runs sensitive automation in production without appropriate guardrails.