- Refactor model handling to support both Claude and Cursor models across various components.
- Introduce `stripProviderPrefix` utility for consistent model ID processing.
- Update `CursorProvider` to utilize `isCursorModel` for model validation.
- Implement model override functionality in GitHub issue validation and enhancement routes.
- Add `useCursorStatusInit` hook to initialize Cursor CLI status on app startup.
- Update UI components to reflect changes in model selection and validation processes.
This update improves the flexibility of AI model usage and enhances user experience by allowing quick model overrides.
- Add PhaseModelConfig interface with 8 configurable phases:
- Quick tasks: enhancement, fileDescription, imageDescription
- Validation: validationModel
- Generation: specGeneration, featureGeneration, backlogPlanning, projectAnalysis
- Add PhaseModelKey type for type-safe access
- Add DEFAULT_PHASE_MODELS with sensible defaults
- Add phaseModels field to GlobalSettings
- Mark legacy enhancementModel/validationModel as deprecated
- Export new types from @automaker/types
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
This commit fixes two issues introduced during prompt customization:
1. **Restored Full Planning Prompts from Main**
- Lite Mode: Added "Silently analyze the codebase first" instruction
- Spec Mode: Restored detailed task format rules, [TASK_START]/[TASK_COMPLETE] markers
- Full Mode: Restored comprehensive SDD format with [PHASE_COMPLETE] markers
- Fixed table structures (Files to Modify, Technical Context, Risks & Mitigations)
- Ensured all critical instructions for Auto Mode functionality are preserved
2. **Fixed Test Suite (774 tests passing)**
- Made getPlanningPromptPrefix() async-aware in all 11 planning tests
- Replaced console.log/error mocks with createLogger mocks (settings-helpers, agent-service)
- Updated test expectations to match restored prompts
- Fixed variable hoisting issue in agent-service mock setup
- Built prompts library to apply changes
The planning prompts now match the detailed, production-ready versions from main
branch, ensuring Auto Mode has all necessary instructions for proper task execution.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
Add comprehensive prompt customization system allowing users to customize
all AI prompts (Auto Mode, Agent Runner, Backlog Plan, Enhancement) through
the Settings UI.
## Features
### Core Customization System
- New TypeScript types for prompt customization with enabled flag
- CustomPrompt interface with value and enabled state
- Prompts preserved even when disabled (no data loss)
- Merged prompt system (custom overrides defaults when enabled)
- Persistent storage in ~/.automaker/settings.json
### Settings UI
- New "Prompt Customization" section in Settings
- 4 tabs: Auto Mode, Agent, Backlog Plan, Enhancement
- Toggle-based editing (read-only default → editable custom)
- Dynamic textarea height based on prompt length (120px-600px)
- Visual state indicators (Custom/Default labels)
### Warning System
- Critical prompt warnings for Backlog Plan (JSON format requirement)
- Field-level warnings when editing critical prompts
- Info banners for Auto Mode planning markers
- Color-coded warnings (blue=info, amber=critical)
### Backend Integration
- Auto Mode service loads prompts from settings
- Agent service loads prompts from settings
- Backlog Plan service loads prompts from settings
- Enhancement endpoint loads prompts from settings
- Settings sync includes promptCustomization field
### Files Changed
- libs/types/src/prompts.ts - Type definitions
- libs/prompts/src/defaults.ts - Default prompt values
- libs/prompts/src/merge.ts - Merge utilities
- apps/ui/src/components/views/settings-view/prompts/ - UI components
- apps/server/src/lib/settings-helpers.ts - getPromptCustomization()
- All service files updated to use customizable prompts
## Technical Details
Prompt storage format:
```json
{
"promptCustomization": {
"autoMode": {
"planningLite": {
"value": "Custom prompt text...",
"enabled": true
}
}
}
}
```
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
- Updated AddFeatureDialog to support both Cursor and Claude profiles, allowing for dynamic model and thinking level selection based on the chosen profile.
- Modified ModelSelector to filter available Cursor models based on global settings and display a warning if the Cursor CLI is not available.
- Enhanced ProfileQuickSelect to handle both profile types and improve selection logic for Cursor profiles.
- Refactored CursorSettingsTab to manage global settings for enabled Cursor models and default model selection, streamlining the configuration process.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Address code review feedback from Gemini Code Assist:
1. Reduce duplication in ClaudeProvider catch block
- Consolidate error creation logic into single path
- Use conditional message building instead of duplicate blocks
- Improves maintainability and follows DRY principle
2. Better separation of concerns in error utilities
- Move default retry-after (60s) logic from extractRetryAfter to classifyError
- extractRetryAfter now only extracts explicit values
- classifyError provides default using nullish coalescing (?? 60)
- Clearer single responsibility for each function
3. Update test to match new behavior
- extractRetryAfter now returns undefined for rate limits without explicit value
- Default value is tested in classifyError tests instead
All 162 tests still passing ✅
Builds successfully with no TypeScript errors ✅
- Add rate_limit error type to ErrorInfo classification
- Implement isRateLimitError() and extractRetryAfter() utilities
- Enhance ClaudeProvider error handling with actionable messages
- Add comprehensive test coverage (8 new tests, 162 total passing)
**Problem:**
When hitting API rate limits, users saw cryptic 'exit code 1' errors
with no explanation or guidance on how to resolve the issue.
**Solution:**
- Detect rate limit errors (429) and extract retry-after duration
- Provide clear, user-friendly error messages with:
* Explanation of what went wrong
* How long to wait before retrying
* Actionable tip to reduce concurrency in auto-mode
- Preserve original error details for debugging
**Changes:**
- libs/types: Add 'rate_limit' type and retryAfter field to ErrorInfo
- libs/utils: Add rate limit detection and extraction logic
- apps/server: Enhance ClaudeProvider with better error messages
- tests: Add 8 new test cases covering rate limit scenarios
**Benefits:**
✅ Clear communication - users understand the problem
✅ Actionable guidance - users know how to fix it
✅ Better debugging - original errors preserved
✅ Type safety - proper TypeScript typing
✅ Comprehensive testing - all edge cases covered
See CHANGELOG_RATE_LIMIT_HANDLING.md for detailed documentation.
- Restricted CORS to localhost origins to prevent remote code execution (RCE) attacks.
- Updated MCP server configuration handling to enforce security warnings when adding or importing servers.
- Introduced a SecurityWarningDialog to inform users about potential risks associated with server commands and configurations.
- Ensured that only serverId is accepted for testing server connections, preventing arbitrary command execution.
These changes improve the overall security posture of the MCP server management and usage.
- Replace HTML checkbox with proper UI Checkbox component
- Add system prompt instructions for AI to check PR changes via gh CLI
- Add PRAnalysis schema field with recommendation (wait_for_merge, pr_needs_work, no_pr)
- Show detailed PR analysis badge in validation dialog
- Hide "Convert to Task" button when PR fix is ready (wait_for_merge)
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add comments section to issue detail panel with lazy loading
- Fetch comments via GraphQL API with pagination (50 at a time)
- Include comments in AI validation analysis when checkbox enabled
- Pass linked PRs info to AI validation for context
- Add "Work in Progress" badge in validation dialog for open PRs
- Add debug logging for validation requests
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add MCPTestService for testing MCP server connections
- Support stdio, SSE, and HTTP transport types
- Implement workaround for SSE headers bug (SDK Issue #436)
- Create API routes for /api/mcp/test and /api/mcp/tools
- Add API client methods for MCP operations
- Create MCPToolsList component with collapsible schema display
- Add Test button to MCP servers section with status indicators
- Add Headers field for HTTP/SSE servers
- Add Environment Variables field for stdio servers
- Fix text overflow in tools list display
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add reusable WSL utilities in @automaker/platform (wsl.ts):
- isWslAvailable() - Check if WSL is available on Windows
- findCliInWsl() - Find CLI tools in WSL, tries multiple distributions
- execInWsl() - Execute commands in WSL
- createWslCommand() - Create spawn-compatible command/args for WSL
- windowsToWslPath/wslToWindowsPath - Path conversion utilities
- getWslDistributions() - List available WSL distributions
- Update CursorProvider to use WSL on Windows:
- Detect cursor-agent in WSL distributions (prioritizes Ubuntu)
- Use full path to wsl.exe for spawn() compatibility
- Pass --cd flag for working directory inside WSL
- Store and use WSL distribution for all commands
- Show "(WSL:Ubuntu) /path" in installation status
- Add 'wsl' to InstallationStatus.method type
- Fix bugs:
- Fix ternary in settings-view.tsx that always returned 'claude'
- Fix findIndex -1 handling in WSL command construction
- Remove 'gpt-5.2' from unknown models test (now valid Cursor model)
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Introduced a new pipeline service to manage custom workflow steps that execute after a feature is marked "In Progress".
- Added API endpoints for configuring, saving, adding, updating, deleting, and reordering pipeline steps.
- Enhanced the UI to support pipeline settings, including a dialog for managing steps and integration with the Kanban board.
- Updated the application state management to handle pipeline configurations per project.
- Implemented dynamic column generation in the Kanban board to display pipeline steps between "In Progress" and "Waiting Approval".
- Added documentation for the new pipeline feature, including usage instructions and configuration details.
This feature allows for a more structured workflow, enabling automated processes such as code reviews and testing after feature implementation.
- Replaced outdated model IDs with new versions, including Claude Sonnet 4.5 and Claude Opus 4.5.
- Added new models such as Gemini 3 Pro, GPT-5.1, and GPT-5.2 with various configurations.
- Enhanced model metadata with descriptions and thinking capabilities for improved clarity and usability.
- Added support for Cursor models in the model resolver, allowing cursor-prefixed models to pass through unchanged.
- Implemented logic to handle bare Cursor model IDs by adding the cursor- prefix.
- Updated logging to provide detailed information on model resolution processes for both Claude and Cursor models.
- Expanded unit tests to cover new Cursor model handling scenarios, ensuring robust validation of model resolution logic.
- Created a comprehensive documentation file detailing the steps to add new Cursor CLI models to AutoMaker.
- Included an overview of the necessary types and configurations, along with a step-by-step guide for model integration.
- Provided examples and a checklist to ensure proper implementation and verification of new models in the UI.
Add Model Context Protocol (MCP) server integration to extend AI agent
capabilities with external tools. This allows users to configure MCP
servers (stdio, SSE, HTTP) in global settings and have agents use them.
Note: MCP servers are currently configured globally. Per-project MCP
server configuration is planned for a future update.
Features:
- New MCP Servers settings section with full CRUD operations
- Import/Export JSON configs (Claude Code format compatible)
- Configurable permission settings:
- Auto-approve MCP tools (bypass permission prompts)
- Unrestricted tools (allow all tools when MCP enabled)
- Refresh button to reload from settings file
Implementation:
- Added MCPServerConfig and MCPToolInfo types
- Added store actions for MCP server management
- Updated claude-provider to use configurable MCP permissions
- Updated sdk-options factory functions for MCP support
- Added settings helpers for loading MCP configs
- Updated AIProfile type to include support for Cursor provider, adding cursorModel and validation logic.
- Enhanced ProfileForm component to handle provider selection and corresponding model configurations for both Claude and Cursor.
- Implemented display functions for model and thinking configurations in ProfileQuickSelect.
- Added default Cursor profiles to the application state.
- Updated UI components to reflect provider-specific settings and validations.
- Marked completion of the AI Profiles Integration phase in the project plan.
- Introduced new types and interfaces for Cursor CLI configuration, authentication status, and event handling.
- Created a comprehensive model definition for Cursor models, including metadata and helper functions.
- Updated existing interfaces to support both Claude and Cursor models in the UI.
- Enhanced the default model configuration to include Cursor's recommended default.
- Updated type exports to include new Cursor-related definitions.
The sandbox configuration was set in createChatOptions() and
createAutoModeOptions(), but was never passed to the ClaudeProvider.
This caused the sandbox to never actually be enabled.
Changes:
- Add sandbox field to ExecuteOptions interface
- Pass sandbox config from AgentService to provider
- Pass sandbox config from AutoModeService to provider
- Forward sandbox config in ClaudeProvider to SDK options
Now the sandbox configuration from settings is properly used.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
Add a global setting to enable/disable sandbox mode for Claude Agent SDK.
This allows users to control sandbox behavior based on their authentication
setup and system compatibility.
Changes:
- Add enableSandboxMode to GlobalSettings (default: true)
- Add sandbox mode checkbox in Claude settings UI
- Wire up setting through app store and settings service
- Update createChatOptions and createAutoModeOptions to use setting
- Add getEnableSandboxModeSetting helper function
- Remove hardcoded sandbox configuration from ClaudeProvider
- Add detailed logging throughout agent execution flow
The sandbox mode requires API key or OAuth token authentication. Users
experiencing issues with CLI-only auth can disable it in settings.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
- Enhanced the configureThrottling function to prevent changes to maxConcurrency while operations are in flight.
- Added comprehensive unit tests for secure-fs throttling and retry logic, ensuring correct behavior and configuration.
- Removed outdated secure-fs test file and replaced it with a new, updated version to improve test coverage.
- Added concurrency limiting using p-limit to prevent ENFILE/EMFILE errors.
- Introduced retry logic with exponential backoff for transient file descriptor errors.
- Enhanced secure-fs with new functions for configuring throttling and monitoring active/pending operations.
- Added unit tests for throttling and retry logic to ensure reliability.
Move directory check before binary file check to handle edge cases
where a directory has a binary file extension (e.g., "images.png/").
Previously, such directories would be incorrectly treated as binary
files instead of being expanded.
Address PR review feedback: replace Promise.all with sequential for...of
loop to avoid exhausting file descriptors when processing directories
with many files.
- Updated API routes to accept an optional settings service for loading the autoLoadClaudeMd setting.
- Introduced a new settings helper utility for retrieving project-specific settings.
- Enhanced feature generation and spec generation processes to utilize the autoLoadClaudeMd setting.
- Refactored relevant route handlers to support the new settings integration across various endpoints.
Previously, when git status reported an untracked directory (e.g., "?? apps/"),
the code would try to read the directory as a file, which failed and showed
"[Unable to read file content]".
Now, when encountering a directory:
- Strip trailing slash from path (git reports dirs as "dirname/")
- Check if path is a directory using stats.isDirectory()
- Recursively list all files inside using listAllFilesInDirectory
- Generate synthetic diffs for each file found
This ensures users see the actual file contents in the diff view instead
of an error placeholder.
- Introduced a new setting to enable automatic loading of CLAUDE.md files from project-specific directories.
- Updated relevant services and components to support the new setting, including the AgentService and AutoModeService.
- Added UI controls for managing the auto-load setting in the settings view.
- Enhanced SDK options to incorporate settingSources for CLAUDE.md loading.
- Updated global and project settings interfaces to include autoLoadClaudeMd property.
- Added onDeleteTask functionality to allow task deletion from both board and graph views.
- Integrated delete options for dependencies in the graph view, enhancing user interaction.
- Updated ancestor context section to clarify the role of parent tasks in task descriptions.
- Improved layout handling in graph view to preserve node positions during updates.
This update enhances task management capabilities and improves user experience in the graph view.
- Updated the `createMarkViewedHandler` to emit an event when a validation is marked as viewed, allowing the UI to update the unviewed count dynamically.
- Modified the `useUnviewedValidations` hook to handle the new event type for decrementing the unviewed validations count.
- Introduced a new event type `issue_validation_viewed` in the issue validation event type definition for better event handling.
- Added a scheduled task to clean up stale validation entries every hour, preventing memory leaks.
- Enhanced the `getAllValidations` function to read validation files in parallel for improved performance.
- Updated the `fetchLinkedPRs` function to use `spawn` for safer execution of GitHub CLI commands, mitigating shell injection risks.
- Modified event handling in the GitHub issues view to utilize the model for validation, ensuring consistency and reducing stale closure issues.
- Introduced a new property in the issue validation event to track the model used for validation.
- Implemented a new function to mark validations as viewed by the user, updating the validation state accordingly.
- Added a new API endpoint for marking validations as viewed, integrated with the existing GitHub routes.
- Enhanced the sidebar to display the count of unviewed validations, providing real-time updates.
- Updated the GitHub issues view to mark validations as viewed when issues are accessed, improving user interaction.
- Introduced a visual indicator for unviewed validations in the issue list, enhancing user awareness of pending validations.
This update introduces a new function, createNewFileDiff, to streamline the generation of synthetic diffs for untracked files. The function reduces code duplication by handling the diff formatting for new files, including directories and large files, improving overall maintainability.
- Introduced CRUD operations for GitHub issue validation results, including storage and retrieval.
- Added new endpoints for checking validation status, stopping validations, and deleting stored validations.
- Enhanced the GitHub routes to support validation management features.
- Updated the UI to display validation results and manage validation states for GitHub issues.
- Integrated event handling for validation progress and completion notifications.
- Added a new endpoint for validating GitHub issues using the Claude SDK.
- Introduced validation schema and logic to handle issue validation requests.
- Updated GitHub routes to include the new validation route.
- Enhanced the UI with a validation dialog and button to trigger issue validation.
- Mapped issue complexity to feature priority for better task management.
- Integrated validation results display in the UI, allowing users to convert validated issues into tasks.
