- Introduced a new `package-lock.json` to manage dependencies.
- Removed obsolete `.automaker/feature_list.json` and replaced it with a new structure under `.automaker/features/{id}/feature.json` for better organization.
- Updated various components to utilize the new features API for managing features, including creation, updates, and deletions.
- Enhanced the UI to reflect changes in feature management, including updates to the sidebar and board view.
- Improved documentation and comments throughout the codebase to clarify the new feature management process.
Add comprehensive worktree management system to enable task isolation and rollback capabilities. This allows users to revert agent changes if they don't satisfy requirements or break functionality.
Key components:
- New WorktreeManager service for branch and worktree operations
- GitDiffPanel component for visualizing changes
- Enhanced UI components with worktree integration
- Auto-mode service enhancements for worktree workflow
Modified files: worktree-manager.js, git-diff-panel.tsx, main.js, preload.js, feature-loader.js, agent-output-modal.tsx, board-view.tsx, kanban-card.tsx, electron.ts, app-store.ts, electron.d.ts
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Sonnet 4 <noreply@anthropic.com>
Resolved conflicts:
- feature_list.json: Merged all features from both branches
- feature-loader.js: Included both model/thinkingLevel and error fields
- board-view.tsx: Merged model/thinkingLevel and error fields, kept currentProject check
- settings-view.tsx: Merged CLI status checks with navigation/scroll code
- app-store.ts: Included both model/thinkingLevel and error fields in Feature interface
Fixed linting errors in settings-view.tsx
- Improved error handling in AutoModeService to log errors and update feature status to "waiting_approval" when an error occurs during feature execution.
- Added error message writing to the context file for better debugging.
- Updated FeatureLoader to include an optional error message parameter when updating feature status.
- Enhanced prompt generation to include detailed context about attached images for better user guidance during feature implementation.
- Modified KanbanCard component to visually indicate features with errors, improving user awareness of issues.
These changes significantly enhance the robustness of the auto mode feature and improve the user experience by providing clearer feedback on feature execution status.
Introduces a comprehensive strategy to safeguard the feature_list.json file from accidental modifications. Key enhancements include:
1. **Prompt-Level Warnings**: Added explicit warnings in agent prompts to prevent direct modifications.
2. **Dedicated MCP Tool**: Implemented the UpdateFeatureStatus tool for safe feature updates.
3. **File-Level Validation & Auto-Backup**: Added validation checks and automatic backups before modifications to prevent data loss.
4. **Tool Access Control**: Restricted agent access to critical tools, ensuring only the designated MCP tool can modify the feature list.
This update significantly reduces the risk of catastrophic data loss and ensures a robust feature management process.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
- Added a new service to detect the installation status of Claude Code CLI, providing users with installation recommendations and commands.
- Integrated CLI detection into the SettingsView to inform users about the CLI status and its benefits for ultrathink tasks.
- Enhanced feature creation and editing dialogs to allow users to select from multiple models (Haiku, Sonnet, Opus) and specify thinking levels (None, Low, Medium, High, Ultrathink).
- Updated the feature executor to utilize the selected model and thinking configuration during task execution, improving flexibility and performance.
This update enhances user experience by providing clearer options for model selection and ensuring optimal performance with the Claude CLI.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Sonnet 4 <noreply@anthropic.com>
Enhance Kanban cards to show comprehensive feature information:
- Add feature summary field to display completed work details
- Show agent context info (model, progress, tasks) on cards
- Display tool usage and token consumption metrics
- Add expandable summary dialog for detailed view
- Update prompts to require summary when marking features complete
- Improve UI for waiting_approval and verified states
Modified: kanban-card.tsx, board-view.tsx, feature-loader.js,
mcp-server-factory.js, prompt-builder.js, app-store.ts
Created: agent-context-parser.ts
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Sonnet 4 <noreply@anthropic.com>
- Added functionality to allow users to send follow-up prompts for features in the waiting_approval status, enabling continued work with additional instructions.
- Implemented a commit feature that allows users to mark waiting_approval features as verified and commit changes directly.
- Updated the UI to include buttons for follow-up and commit actions on Kanban cards and integrated dialogs for user interaction.
- Enhanced the feature loader and executor to handle the new status and actions appropriately.
This update improves the workflow for managing features that require manual review and enhances user experience in the auto mode.
- Added a new keyboard shortcut 'P' to open the project picker dropdown.
- Implemented functionality to select projects using number keys, allowing users to quickly switch between projects.
- Updated the feature list to include a new feature for project selection via keyboard shortcuts.
- Removed obsolete coding_prompt.md and added initializer_prompt.md for better session management.
- Introduced context management for features, enabling reading, writing, and deleting context files.
- Updated package dependencies to include @radix-ui/react-checkbox for enhanced UI components.
This commit enhances user experience by streamlining project selection and improves the overall feature management process.
🤖 Generated with Claude Code
