red64-cli 0.1.0 → 0.3.0

package/framework/agents/claude/.claude/agents/red64/spec-impl.md

@@ -46,6 +46,16 @@ Execute implementation tasks for feature using Test-Driven Development.
 **Read all necessary context**:
 - `.red64/specs/{feature}/spec.json`, `requirements.md`, `design.md`, `tasks.md`
 - **Entire `.red64/steering/` directory** for complete project memory
+- **`.red64/steering/feedback.md`** - CRITICAL: Contains project-specific test, lint, and dev server commands
+
+**Parse feedback.md for commands**:
+Extract these values from the YAML blocks in `feedback.md`:
+- `test`: Primary test command (e.g., `pnpm test:run`, `uv run pytest`)
+- `lint`: Primary lint command (e.g., `pnpm lint`, `uv run ruff check .`)
+- `dev_server`: Dev server command (e.g., `pnpm dev`, `bin/rails server`)
+- `dev_port`: Dev server port (e.g., `3000`, `5173`, `8000`)
+- `dev_url`: Dev server base URL (e.g., `http://localhost:3000`)
+- `ui_verification_enabled`: Whether to use agent-browser for UI tasks
 
 **Validate approvals**:
 - Verify tasks are approved in spec.json (stop if not, see Safety & Fallback)
@@ -81,17 +91,136 @@ For each selected task, follow Kent Beck's TDD cycle:
    - No regressions in existing functionality
    - Code coverage maintained or improved
 
+5. **FEEDBACK LOOP - Self-Correction**:
+   After each implementation change, run the feedback loop using commands from `feedback.md`:
+
+   **Testing Feedback** (MANDATORY):
+   ```bash
+   # Use the 'test' command from .red64/steering/feedback.md
+   # Examples by stack:
+   #   - React/Next.js: pnpm test:run
+   #   - Python: uv run pytest
+   #   - Rails: bin/rails test
+   {test_command_from_feedback_md}
+   ```
+   - Parse test output for failures
+   - If tests fail: Fix issues before proceeding
+   - Never mark implementation complete with failing tests
+
+   **Linting Feedback** (when available):
+   ```bash
+   # Use the 'lint' command from .red64/steering/feedback.md
+   # Examples by stack:
+   #   - React/Next.js: pnpm lint
+   #   - Python: uv run ruff check .
+   #   - Rails: bundle exec rubocop
+   {lint_command_from_feedback_md}
+   ```
+   - Fix critical/error-level issues
+   - Warnings are advisory (log but proceed)
+
+   **UI Verification** (for UI-related tasks - see UI Verification Protocol below):
+   - Check `ui_verification_enabled` in feedback.md (skip if false)
+   - Required when task involves: components, pages, layouts, styles, visual elements
+   - Use agent-browser for screenshot capture and accessibility analysis
+
 **Note**: Do NOT update task checkboxes in tasks.md. The orchestrator handles task completion tracking.
 
+## UI Verification Protocol
+
+When implementing UI-related tasks (components, pages, layouts, styles, visual elements), use **agent-browser** for verification:
+
+### When to Use UI Verification
+- Task creates or modifies React/Vue/Angular components
+- Task involves styling changes (CSS, Tailwind, styled-components)
+- Task mentions visual elements, layouts, or design specs
+- Design.md includes mockups, wireframes, or visual references
+
+### UI Verification Steps
+
+1. **Start Development Server** (if not running):
+   ```bash
+   # Use 'dev_server' command from .red64/steering/feedback.md
+   # Examples by stack:
+   #   - React: pnpm dev (port 5173)
+   #   - Next.js: pnpm dev (port 3000)
+   #   - Python/FastAPI: uv run uvicorn src.app.main:app --reload (port 8000)
+   #   - Rails: bin/rails server (port 3000)
+   {dev_server_command_from_feedback_md} &>/dev/null &
+   # Wait for server to be ready
+   sleep 5
+   ```
+
+2. **Navigate to Feature**:
+   ```bash
+   # Use 'dev_url' from .red64/steering/feedback.md as base URL
+   agent-browser goto {dev_url_from_feedback_md}/path/to/feature
+   ```
+
+3. **Capture Screenshot**:
+   ```bash
+   agent-browser screenshot --full-page /tmp/ui-capture-$(date +%s).png
+   ```
+
+4. **Get Accessibility Tree** (for structural verification):
+   ```bash
+   agent-browser snapshot > /tmp/accessibility-tree.json
+   ```
+
+5. **Analyze and Compare**:
+   - Read the captured screenshot
+   - Compare against design specifications in design.md
+   - Check accessibility tree for proper element structure
+   - Verify interactive elements are accessible
+
+6. **Fix Visual Discrepancies**:
+   - If UI doesn't match design spec: fix code and repeat steps 2-5
+   - If accessibility issues found: add proper ARIA labels, semantic HTML
+   - Continue until UI matches specification
+
+### Agent-Browser Commands Reference
+
+| Command | Purpose |
+|---------|---------|
+| `agent-browser goto <url>` | Navigate to URL |
+| `agent-browser screenshot [--full-page] <path>` | Capture screenshot |
+| `agent-browser snapshot` | Get AI-optimized accessibility tree |
+| `agent-browser click <element-ref>` | Click on element |
+| `agent-browser type <element-ref> <text>` | Type into input |
+| `agent-browser scroll <direction>` | Scroll page |
+
+### UI Verification Loop
+
+```
+┌─────────────────────────────────────────────────────────┐
+│  1. Implement UI code changes                           │
+│  2. Start dev server (if not running)                   │
+│  3. Navigate: agent-browser goto <url>                  │
+│  4. Capture: agent-browser screenshot /tmp/ui.png       │
+│  5. Analyze screenshot vs. design spec                  │
+│  6. If visual mismatch:                                 │
+│     - Identify discrepancies                            │
+│     - Fix code                                          │
+│     - Goto step 3                                       │
+│  7. Capture accessibility tree for structure check      │
+│  8. Fix accessibility issues if found                   │
+│  9. Run tests to ensure no regressions                  │
+│  10. Mark implementation complete                       │
+└─────────────────────────────────────────────────────────┘
+```
+
 ## Critical Constraints
 - **TDD Mandatory**: Tests MUST be written before implementation code
 - **Task Scope**: Implement only what the specific task requires
 - **Test Coverage**: All new code must have tests
 - **No Regressions**: Existing tests must continue to pass
 - **Design Alignment**: Implementation must follow design.md specifications
+- **Feedback Loop**: Run tests after EVERY implementation change; fix failures before proceeding
+- **UI Verification**: For UI-related tasks, use agent-browser to capture and verify visual output
 
 ## Tool Guidance
-- **Read first**: Load all context before implementation
+- **Read feedback.md first**: Parse `.red64/steering/feedback.md` to get project-specific test, lint, and dev server commands
+- **Read all context**: Load specs and steering before implementation
 - **Test first**: Write tests before code
 - Use **WebSearch/WebFetch** for library documentation when needed
 
@@ -117,4 +246,4 @@ Provide brief summary in the language specified in spec.json:
 - **Action**: Debug and fix, then re-run
 
 **Note**: You execute tasks autonomously. Return final report only when complete.
-think
+think

package/framework/agents/claude/.claude/agents/red64/validate-gap.md

@@ -1,7 +1,7 @@
 ---
 name: validate-gap-agent
 description: Analyze implementation gap between requirements and existing codebase
-tools: Read, Grep, Glob, WebSearch, WebFetch
+tools: Read, Write, Edit, Grep, Glob, WebSearch, WebFetch
 model: inherit
 color: yellow
 ---
@@ -49,6 +49,7 @@ Analyze implementation gap for feature based on approved requirements and existi
 
 2. **Read Analysis Guidelines**:
    - Read `.red64/settings/rules/gap-analysis.md` for comprehensive analysis framework
+   - Read `.red64/settings/templates/specs/gap-analysis.md` for document structure
 
 3. **Execute Gap Analysis**:
    - Follow gap-analysis.md framework for thorough investigation
@@ -57,11 +58,15 @@ Analyze implementation gap for feature based on approved requirements and existi
    - Evaluate multiple implementation approaches (extend/new/hybrid)
    - Use language specified in spec.json for output
 
-4. **Generate Analysis Document**:
-   - Create comprehensive gap analysis following the output guidelines in gap-analysis.md
+4. **Write Analysis Document**:
+   - **Write gap analysis to `.red64/specs/{feature}/gap-analysis.md`** using the template structure
+   - Include: Requirement-to-Asset Map, Options A/B/C, Effort/Risk assessments, Recommendations
    - Present multiple viable options with trade-offs
    - Flag areas requiring further research
 
+5. **Update Metadata**:
+   - Edit `.red64/specs/{feature}/spec.json` to set `phase: "gap-analyzed"` if needed
+
 ## Important Constraints
 - **Information over Decisions**: Provide analysis and options, not final implementation choices
 - **Multiple Options**: Present viable alternatives when applicable
@@ -69,22 +74,23 @@ Analyze implementation gap for feature based on approved requirements and existi
 - **Explicit Gaps**: Clearly flag areas needing research or investigation
 
 ## Tool Guidance
-- **Read first**: Load all context (spec, steering, rules) before analysis
+- **Read first**: Load all context (spec, steering, rules, templates) before analysis
 - **Grep extensively**: Search codebase for patterns, conventions, and integration points
 - **WebSearch/WebFetch**: Research external dependencies and best practices when needed
-- **Write last**: Generate analysis only after complete investigation
+- **Write last**: Write gap-analysis.md only after complete investigation
 
 ## Output Description
 Provide output in the language specified in spec.json with:
 
 1. **Analysis Summary**: Brief overview (3-5 bullets) of scope, challenges, and recommendations
-2. **Document Status**: Confirm analysis approach used
+2. **Document Status**: Confirm gap-analysis.md was created at `.red64/specs/{feature}/gap-analysis.md`
 3. **Next Steps**: Guide user on proceeding to design phase
 
 **Format Requirements**:
 - Use Markdown headings for clarity
 - Keep summary concise (under 300 words)
-- Detailed analysis follows gap-analysis.md output guidelines
+- Include the full path to the generated gap-analysis.md file
+- Detailed analysis is in the generated document
 
 ## Safety & Fallback
 

package/framework/agents/claude/.claude/commands/red64/spec-impl.md

@@ -25,6 +25,19 @@ If validation fails, inform user to complete tasks generation first.
 - If `$2` provided: Parse task numbers (e.g., "1.1", "1,2,3")
 - Otherwise: Read `.red64/specs/$1/tasks.md` and find all unchecked tasks (`- [ ]`)
 
+## Detect UI-Related Tasks
+
+Before invoking the subagent, analyze the tasks to determine if they involve UI work:
+
+**UI Detection Keywords** (in task description or design.md):
+- Component, page, layout, view, screen
+- Style, CSS, Tailwind, styled-components
+- Button, form, input, modal, dialog
+- Visual, UI, UX, design, mockup
+- Frontend, client-side, browser
+
+If UI-related keywords detected, set `UI Mode: enabled` in the prompt.
+
 ## Invoke Subagent
 
 Delegate TDD implementation to spec-tdd-impl-agent:
@@ -45,6 +58,17 @@ File patterns to read:
 - .red64/steering/*.md
 
 TDD Mode: strict (test-first)
+UI Mode: {enabled if UI-related, disabled otherwise}
+
+## Feedback Requirements
+After EVERY implementation change:
+1. Run project tests: Use command from .red64/steering/feedback.md
+2. Fix any test failures before proceeding
+3. If UI Mode enabled: Use agent-browser for visual verification
+   - Start dev server if not running
+   - Capture screenshots of implemented UI
+   - Compare against design specifications
+   - Fix visual discrepancies before marking complete
 """
 )
 ```

package/framework/agents/claude/.claude/commands/red64/validate-gap.md

@@ -35,6 +35,10 @@ File patterns to read:
 - .red64/specs/$1/requirements.md
 - .red64/steering/*.md
 - .red64/settings/rules/gap-analysis.md
+- .red64/settings/templates/specs/gap-analysis.md
+
+Output:
+- Write gap analysis document to: .red64/specs/$1/gap-analysis.md
 """
 )
 ```

package/framework/agents/codex/.codex/agents/red64/spec-impl.md

@@ -46,6 +46,16 @@ Execute implementation tasks for feature using Test-Driven Development.
 **Read all necessary context**:
 - `.red64/specs/{feature}/spec.json`, `requirements.md`, `design.md`, `tasks.md`
 - **Entire `.red64/steering/` directory** for complete project memory
+- **`.red64/steering/feedback.md`** - CRITICAL: Contains project-specific test, lint, and dev server commands
+
+**Parse feedback.md for commands**:
+Extract these values from the YAML blocks in `feedback.md`:
+- `test`: Primary test command (e.g., `pnpm test:run`, `uv run pytest`)
+- `lint`: Primary lint command (e.g., `pnpm lint`, `uv run ruff check .`)
+- `dev_server`: Dev server command (e.g., `pnpm dev`, `bin/rails server`)
+- `dev_port`: Dev server port (e.g., `3000`, `5173`, `8000`)
+- `dev_url`: Dev server base URL (e.g., `http://localhost:3000`)
+- `ui_verification_enabled`: Whether to use agent-browser for UI tasks
 
 **Validate approvals**:
 - Verify tasks are approved in spec.json (stop if not, see Safety & Fallback)
@@ -81,17 +91,136 @@ For each selected task, follow Kent Beck's TDD cycle:
    - No regressions in existing functionality
    - Code coverage maintained or improved
 
+5. **FEEDBACK LOOP - Self-Correction**:
+   After each implementation change, run the feedback loop using commands from `feedback.md`:
+
+   **Testing Feedback** (MANDATORY):
+   ```bash
+   # Use the 'test' command from .red64/steering/feedback.md
+   # Examples by stack:
+   #   - React/Next.js: pnpm test:run
+   #   - Python: uv run pytest
+   #   - Rails: bin/rails test
+   {test_command_from_feedback_md}
+   ```
+   - Parse test output for failures
+   - If tests fail: Fix issues before proceeding
+   - Never mark implementation complete with failing tests
+
+   **Linting Feedback** (when available):
+   ```bash
+   # Use the 'lint' command from .red64/steering/feedback.md
+   # Examples by stack:
+   #   - React/Next.js: pnpm lint
+   #   - Python: uv run ruff check .
+   #   - Rails: bundle exec rubocop
+   {lint_command_from_feedback_md}
+   ```
+   - Fix critical/error-level issues
+   - Warnings are advisory (log but proceed)
+
+   **UI Verification** (for UI-related tasks - see UI Verification Protocol below):
+   - Check `ui_verification_enabled` in feedback.md (skip if false)
+   - Required when task involves: components, pages, layouts, styles, visual elements
+   - Use agent-browser for screenshot capture and accessibility analysis
+
 **Note**: Do NOT update task checkboxes in tasks.md. The orchestrator handles task completion tracking.
 
+## UI Verification Protocol
+
+When implementing UI-related tasks (components, pages, layouts, styles, visual elements), use **agent-browser** for verification:
+
+### When to Use UI Verification
+- Task creates or modifies React/Vue/Angular components
+- Task involves styling changes (CSS, Tailwind, styled-components)
+- Task mentions visual elements, layouts, or design specs
+- Design.md includes mockups, wireframes, or visual references
+
+### UI Verification Steps
+
+1. **Start Development Server** (if not running):
+   ```bash
+   # Use 'dev_server' command from .red64/steering/feedback.md
+   # Examples by stack:
+   #   - React: pnpm dev (port 5173)
+   #   - Next.js: pnpm dev (port 3000)
+   #   - Python/FastAPI: uv run uvicorn src.app.main:app --reload (port 8000)
+   #   - Rails: bin/rails server (port 3000)
+   {dev_server_command_from_feedback_md} &>/dev/null &
+   # Wait for server to be ready
+   sleep 5
+   ```
+
+2. **Navigate to Feature**:
+   ```bash
+   # Use 'dev_url' from .red64/steering/feedback.md as base URL
+   agent-browser goto {dev_url_from_feedback_md}/path/to/feature
+   ```
+
+3. **Capture Screenshot**:
+   ```bash
+   agent-browser screenshot --full-page /tmp/ui-capture-$(date +%s).png
+   ```
+
+4. **Get Accessibility Tree** (for structural verification):
+   ```bash
+   agent-browser snapshot > /tmp/accessibility-tree.json
+   ```
+
+5. **Analyze and Compare**:
+   - Read the captured screenshot
+   - Compare against design specifications in design.md
+   - Check accessibility tree for proper element structure
+   - Verify interactive elements are accessible
+
+6. **Fix Visual Discrepancies**:
+   - If UI doesn't match design spec: fix code and repeat steps 2-5
+   - If accessibility issues found: add proper ARIA labels, semantic HTML
+   - Continue until UI matches specification
+
+### Agent-Browser Commands Reference
+
+| Command | Purpose |
+|---------|---------|
+| `agent-browser goto <url>` | Navigate to URL |
+| `agent-browser screenshot [--full-page] <path>` | Capture screenshot |
+| `agent-browser snapshot` | Get AI-optimized accessibility tree |
+| `agent-browser click <element-ref>` | Click on element |
+| `agent-browser type <element-ref> <text>` | Type into input |
+| `agent-browser scroll <direction>` | Scroll page |
+
+### UI Verification Loop
+
+```
+┌─────────────────────────────────────────────────────────┐
+│  1. Implement UI code changes                           │
+│  2. Start dev server (if not running)                   │
+│  3. Navigate: agent-browser goto <url>                  │
+│  4. Capture: agent-browser screenshot /tmp/ui.png       │
+│  5. Analyze screenshot vs. design spec                  │
+│  6. If visual mismatch:                                 │
+│     - Identify discrepancies                            │
+│     - Fix code                                          │
+│     - Goto step 3                                       │
+│  7. Capture accessibility tree for structure check      │
+│  8. Fix accessibility issues if found                   │
+│  9. Run tests to ensure no regressions                  │
+│  10. Mark implementation complete                       │
+└─────────────────────────────────────────────────────────┘
+```
+
 ## Critical Constraints
 - **TDD Mandatory**: Tests MUST be written before implementation code
 - **Task Scope**: Implement only what the specific task requires
 - **Test Coverage**: All new code must have tests
 - **No Regressions**: Existing tests must continue to pass
 - **Design Alignment**: Implementation must follow design.md specifications
+- **Feedback Loop**: Run tests after EVERY implementation change; fix failures before proceeding
+- **UI Verification**: For UI-related tasks, use agent-browser to capture and verify visual output
 
 ## Tool Guidance
-- **Read first**: Load all context before implementation
+- **Read feedback.md first**: Parse `.red64/steering/feedback.md` to get project-specific test, lint, and dev server commands
+- **Read all context**: Load specs and steering before implementation
 - **Test first**: Write tests before code
 - Use **WebSearch/WebFetch** for library documentation when needed
 
@@ -117,4 +246,4 @@ Provide brief summary in the language specified in spec.json:
 - **Action**: Debug and fix, then re-run
 
 **Note**: You execute tasks autonomously. Return final report only when complete.
-think
+think

package/framework/agents/codex/.codex/agents/red64/validate-gap.md

@@ -1,7 +1,7 @@
 ---
 name: validate-gap-agent
 description: Analyze implementation gap between requirements and existing codebase
-tools: Read, Grep, Glob, WebSearch, WebFetch
+tools: Read, Write, Edit, Grep, Glob, WebSearch, WebFetch
 model: inherit
 color: yellow
 ---
@@ -49,6 +49,7 @@ Analyze implementation gap for feature based on approved requirements and existi
 
 2. **Read Analysis Guidelines**:
    - Read `.red64/settings/rules/gap-analysis.md` for comprehensive analysis framework
+   - Read `.red64/settings/templates/specs/gap-analysis.md` for document structure
 
 3. **Execute Gap Analysis**:
    - Follow gap-analysis.md framework for thorough investigation
@@ -57,11 +58,15 @@ Analyze implementation gap for feature based on approved requirements and existi
    - Evaluate multiple implementation approaches (extend/new/hybrid)
    - Use language specified in spec.json for output
 
-4. **Generate Analysis Document**:
-   - Create comprehensive gap analysis following the output guidelines in gap-analysis.md
+4. **Write Analysis Document**:
+   - **Write gap analysis to `.red64/specs/{feature}/gap-analysis.md`** using the template structure
+   - Include: Requirement-to-Asset Map, Options A/B/C, Effort/Risk assessments, Recommendations
    - Present multiple viable options with trade-offs
    - Flag areas requiring further research
 
+5. **Update Metadata**:
+   - Edit `.red64/specs/{feature}/spec.json` to set `phase: "gap-analyzed"` if needed
+
 ## Important Constraints
 - **Information over Decisions**: Provide analysis and options, not final implementation choices
 - **Multiple Options**: Present viable alternatives when applicable
@@ -69,22 +74,23 @@ Analyze implementation gap for feature based on approved requirements and existi
 - **Explicit Gaps**: Clearly flag areas needing research or investigation
 
 ## Tool Guidance
-- **Read first**: Load all context (spec, steering, rules) before analysis
+- **Read first**: Load all context (spec, steering, rules, templates) before analysis
 - **Grep extensively**: Search codebase for patterns, conventions, and integration points
 - **WebSearch/WebFetch**: Research external dependencies and best practices when needed
-- **Write last**: Generate analysis only after complete investigation
+- **Write last**: Write gap-analysis.md only after complete investigation
 
 ## Output Description
 Provide output in the language specified in spec.json with:
 
 1. **Analysis Summary**: Brief overview (3-5 bullets) of scope, challenges, and recommendations
-2. **Document Status**: Confirm analysis approach used
+2. **Document Status**: Confirm gap-analysis.md was created at `.red64/specs/{feature}/gap-analysis.md`
 3. **Next Steps**: Guide user on proceeding to design phase
 
 **Format Requirements**:
 - Use Markdown headings for clarity
 - Keep summary concise (under 300 words)
-- Detailed analysis follows gap-analysis.md output guidelines
+- Include the full path to the generated gap-analysis.md file
+- Detailed analysis is in the generated document
 
 ## Safety & Fallback
 

package/framework/agents/codex/.codex/commands/red64/spec-impl.md

@@ -25,6 +25,19 @@ If validation fails, inform user to complete tasks generation first.
 - If `$2` provided: Parse task numbers (e.g., "1.1", "1,2,3")
 - Otherwise: Read `.red64/specs/$1/tasks.md` and find all unchecked tasks (`- [ ]`)
 
+## Detect UI-Related Tasks
+
+Before invoking the subagent, analyze the tasks to determine if they involve UI work:
+
+**UI Detection Keywords** (in task description or design.md):
+- Component, page, layout, view, screen
+- Style, CSS, Tailwind, styled-components
+- Button, form, input, modal, dialog
+- Visual, UI, UX, design, mockup
+- Frontend, client-side, browser
+
+If UI-related keywords detected, set `UI Mode: enabled` in the prompt.
+
 ## Invoke Subagent
 
 Delegate TDD implementation to spec-tdd-impl-agent:
@@ -45,6 +58,17 @@ File patterns to read:
 - .red64/steering/*.md
 
 TDD Mode: strict (test-first)
+UI Mode: {enabled if UI-related, disabled otherwise}
+
+## Feedback Requirements
+After EVERY implementation change:
+1. Run project tests: Use command from .red64/steering/feedback.md
+2. Fix any test failures before proceeding
+3. If UI Mode enabled: Use agent-browser for visual verification
+   - Start dev server if not running
+   - Capture screenshots of implemented UI
+   - Compare against design specifications
+   - Fix visual discrepancies before marking complete
 """
 )
 ```

package/framework/agents/codex/.codex/commands/red64/validate-gap.md

@@ -35,6 +35,10 @@ File patterns to read:
 - .red64/specs/$1/requirements.md
 - .red64/steering/*.md
 - .red64/settings/rules/gap-analysis.md
+- .red64/settings/templates/specs/gap-analysis.md
+
+Output:
+- Write gap analysis document to: .red64/specs/$1/gap-analysis.md
 """
 )
 ```

package/framework/stacks/generic/feedback.md

@@ -0,0 +1,80 @@
+# Feedback Configuration
+
+Project-specific commands for automated feedback during implementation.
+
+---
+
+## Test Commands
+
+Commands to run tests during implementation. The agent will use these to verify code changes.
+
+```yaml
+# Primary test command (REQUIRED)
+test: npm test
+
+# Test with coverage report
+test_coverage: npm test -- --coverage
+
+# Run specific test file (use {file} as placeholder)
+test_file: npm test -- {file}
+```
+
+---
+
+## Linting Commands
+
+Commands for code quality checks.
+
+```yaml
+# Primary lint command
+lint: npm run lint
+
+# Lint with auto-fix
+lint_fix: npm run lint -- --fix
+
+# Type checking (if applicable)
+type_check: npm run type-check
+```
+
+---
+
+## Development Server
+
+Commands for starting the development server (required for UI verification).
+
+```yaml
+# Start dev server
+dev_server: npm run dev
+
+# Dev server port
+dev_port: 3000
+
+# Dev server base URL
+dev_url: http://localhost:3000
+```
+
+---
+
+## UI Verification
+
+Settings for agent-browser UI verification.
+
+```yaml
+# Enable UI verification for this project
+ui_verification_enabled: true
+
+# Default wait time after navigation (milliseconds)
+navigation_wait: 3000
+
+# Screenshot directory
+screenshot_dir: /tmp/ui-captures
+```
+
+---
+
+## Notes
+
+- Update these commands to match your project's setup
+- The agent reads this file to determine how to run tests and verify UI
+- If a command doesn't apply, leave it empty or remove the line
+- All commands should be runnable from the project root directory