feat: scaffold DojoWatch as Claude Code plugin + CI scripts (Phase 1)

AI-native visual regression testing engine — captures screenshots with
Playwright, pre-filters with pixelmatch, and uses Claude (locally) or
Gemini (in CI) as the diff engine.

Plugin components:
- 5 slash commands: /vr-init, /vr-check, /vr-approve, /vr-report, /vr-watch
- regression-analyzer agent for root-cause tracing
- visual-regression skill with classification schema + analysis prompt

Core scripts:
- capture.ts: Playwright capture engine with stabilization and masking
- baseline.ts: SHA-256 hashing, baseline promotion
- config.ts: .dojowatch/config.json loader with defaults
- route-map.ts: source file → route mapping with git-aware scope resolution
- types.ts: shared TypeScript types

Also includes CI workflow, issue templates, config example, and
GitHub Actions template for consumer projects.

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

Author: 2 people

.claude-plugin/plugin.json

@@ -0,0 +1,19 @@
+{
+  "name": "dojowatch",
+  "version": "0.1.0",
+  "description": "AI-native visual regression testing — capture, diff, and analyze UI changes with LLM vision",
+  "author": {
+    "name": "Dojo Coding LLC",
+    "url": "https://github.com/DojoCodingLabs"
+  },
+  "repository": "https://github.com/DojoCodingLabs/dojowatch",
+  "license": "MIT",
+  "keywords": [
+    "visual-regression",
+    "testing",
+    "playwright",
+    "ai",
+    "screenshot",
+    "diff"
+  ]
+}

.github/ISSUE_TEMPLATE/bug_report.md

@@ -0,0 +1,37 @@
+---
+name: Bug Report
+about: Report a bug in DojoWatch
+title: "[Bug] "
+labels: bug
+---
+
+## Description
+
+A clear description of the bug.
+
+## Steps to reproduce
+
+1.
+2.
+3.
+
+## Expected behavior
+
+What should happen.
+
+## Actual behavior
+
+What actually happens.
+
+## Environment
+
+- DojoWatch version:
+- Node.js version:
+- OS:
+- Framework (Next.js, Vite, etc.):
+- Using Claude Code plugin: yes/no
+- Using CI (GitHub Actions): yes/no
+
+## Screenshots
+
+If applicable, add screenshots.

.github/ISSUE_TEMPLATE/feature_request.md

@@ -0,0 +1,22 @@
+---
+name: Feature Request
+about: Suggest a new feature for DojoWatch
+title: "[Feature] "
+labels: enhancement
+---
+
+## Problem
+
+What problem does this feature solve?
+
+## Proposed solution
+
+How should it work?
+
+## Alternatives considered
+
+Any alternative approaches you've thought about.
+
+## Additional context
+
+Any other context, screenshots, or references.

.github/workflows/ci.yml

@@ -0,0 +1,26 @@
+name: CI
+
+on:
+  push:
+    branches: [main]
+  pull_request:
+    branches: [main]
+
+jobs:
+  check:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+
+      - uses: actions/setup-node@v4
+        with:
+          node-version: 24
+          cache: npm
+
+      - run: npm ci
+
+      - name: Type check
+        run: npm run typecheck
+
+      - name: Test
+        run: npm run test:run

.gitignore

@@ -0,0 +1,31 @@
+# Dependencies
+node_modules/
+
+# Build output
+dist/
+
+# DojoWatch runtime artifacts (captures and diffs are ephemeral)
+.dojowatch/captures/
+.dojowatch/diffs/
+
+# Environment
+.env
+.env.local
+.env*.local
+
+# IDE
+.vscode/
+.idea/
+*.swp
+*.swo
+*~
+
+# OS
+.DS_Store
+Thumbs.db
+
+# TypeScript
+*.tsbuildinfo
+
+# Test coverage
+coverage/

CHANGELOG.md

@@ -0,0 +1,16 @@
+# Changelog
+
+## [0.1.0] — Unreleased
+
+### Added
+- Initial project structure as Claude Code plugin + standalone scripts
+- **Capture engine** (`scripts/capture.ts`): Playwright-based screenshot capture with configurable routes and viewports
+- **Stabilization** (`scripts/stabilize.ts`): Animation freeze, network idle wait, font loading, element masking
+- **Baseline management** (`scripts/baseline.ts`): SHA-256 hashing, promote captures to baselines
+- **Config loader** (`scripts/config.ts`): `.dojowatch/config.json` with sensible defaults
+- **Route map** (`scripts/route-map.ts`): Source file → route mapping with git-aware scope resolution
+- **Slash commands**: `/vr-init`, `/vr-check`, `/vr-approve`, `/vr-report`, `/vr-watch` (stub)
+- **Agent**: `regression-analyzer` for deep-diving into regression root causes
+- **Skill**: `visual-regression` with classification schema and analysis prompt references
+- **Templates**: Example config, GitHub Actions workflow
+- **CI**: GitHub Actions workflow for linting, testing, type checking

CLAUDE.md

@@ -0,0 +1,47 @@
+# DojoWatch
+
+AI-native visual regression testing engine — Claude Code plugin + GitHub Actions CI.
+
+## Architecture
+
+- **Plugin-first**: Slash commands are the primary local interface. Claude Code IS the local AI engine.
+- **Scripts in `scripts/`**: Standalone TypeScript, runnable via `npx tsx`. Used by both slash commands and CI.
+- **Gemini for CI**: `scripts/analyze-gemini.ts` handles batch analysis in GitHub Actions.
+- **No npm publish**: This is a Claude Code plugin, not a published package.
+
+## Conventions
+
+- TypeScript strict mode, ESM only (`"type": "module"`)
+- All shared types in `scripts/types.ts`
+- Pre-filter has a **zero false-negative guarantee**: only byte-identical screenshots skip analysis
+- Scripts must be runnable standalone (for CI) AND callable from slash commands via Bash
+- Use `picocolors` for terminal output, never `chalk`
+- Use `execFileSync` not `execSync` — avoid shell injection
+- Test fixtures in `tests/fixtures/`
+
+## File layout
+
+| Directory | Purpose |
+|-----------|---------|
+| `commands/` | Claude Code slash commands (markdown with YAML frontmatter) |
+| `agents/` | Claude Code agents (markdown with YAML frontmatter) |
+| `skills/` | Claude Code skills (`SKILL.md` + `references/`) |
+| `scripts/` | Core TypeScript scripts (capture, prefilter, analyze, etc.) |
+| `templates/` | User-facing templates (config example, GitHub Actions) |
+| `tests/` | Vitest tests + PNG fixtures |
+
+## Script naming
+
+Scripts double as both importable modules and CLI entrypoints:
+- Export functions for programmatic use
+- Include `if (isDirectRun)` block for CLI execution
+- Accept args via `process.argv` when run directly
+
+## Classification schema
+
+Visual diffs are classified as:
+- **REGRESSION**: Unintended visual change (bug). Severity: high/medium/low.
+- **INTENTIONAL**: Deliberate change (feature, design update).
+- **NOISE**: Insignificant rendering variance (anti-aliasing, sub-pixel).
+
+See `skills/visual-regression/references/classification-schema.md` for full criteria.

CONTRIBUTING.md

@@ -0,0 +1,94 @@
+# Contributing to DojoWatch
+
+Thanks for your interest in contributing to DojoWatch!
+
+## Development setup
+
+1. Clone the repo:
+   ```bash
+   git clone https://github.com/DojoCodingLabs/dojowatch.git
+   cd dojowatch
+   ```
+
+2. Install dependencies:
+   ```bash
+   npm install
+   ```
+
+3. Install Playwright browsers:
+   ```bash
+   npx playwright install chromium
+   ```
+
+4. Run tests:
+   ```bash
+   npm test
+   ```
+
+5. Test the plugin locally:
+   ```bash
+   claude --plugin-dir ./
+   ```
+
+## Project structure
+
+- `commands/` — Slash commands for Claude Code (markdown)
+- `agents/` — Claude Code agents (markdown)
+- `skills/` — Auto-activating skills with references
+- `scripts/` — Core TypeScript modules (capture, prefilter, baseline, etc.)
+- `tests/` — Vitest tests with PNG fixtures
+- `templates/` — User-facing config and CI templates
+
+## Writing scripts
+
+Scripts serve dual purposes — importable modules AND standalone CLI tools:
+
+```typescript
+// Export functions for programmatic use
+export function myFunction() { ... }
+
+// CLI entrypoint when run directly
+const isDirectRun = process.argv[1]?.endsWith("my-script.ts");
+if (isDirectRun) {
+  main();
+}
+```
+
+## Writing slash commands
+
+Commands are markdown files in `commands/` with YAML frontmatter:
+
+```markdown
+---
+description: Short description shown in /help
+argument-hint: "[optional args]"
+---
+
+# Command Name
+
+Instructions for Claude on how to execute this command...
+```
+
+Commands instruct Claude what to do — they run scripts via Bash and use Claude's multimodal capabilities for analysis.
+
+## Testing
+
+```bash
+npm test          # Watch mode
+npm run test:run  # Single run
+npm run typecheck # Type checking
+npm run lint      # Linting
+```
+
+Test fixtures (PNG pairs) live in `tests/fixtures/`. When adding new test cases, include PNG pairs that demonstrate the specific scenario.
+
+## Pull requests
+
+1. Create a branch from `main`
+2. Make your changes
+3. Ensure `npm run typecheck && npm run test:run` passes
+4. Submit a PR with a clear description of what changed and why
+
+## Code of conduct
+
+Be kind, be constructive, be collaborative.