rafcode 1.0.0

package/RAF/001-raf-task-improvements/plans/003-stash-changes-on-failure.md

@@ -0,0 +1,37 @@
+# Task: Stash Changes on Complete Failure
+
+## Objective
+Automatically git stash uncommitted changes when a task completely fails (no more retries planned).
+
+## Context
+When a task fails after all retry attempts, any changes Claude made may be left in the working directory in an incomplete state. This task adds automatic stashing of these changes to preserve them while leaving the working directory clean.
+
+## Requirements
+- Stash changes only when task completely fails (all retries exhausted, no more attempts)
+- Do NOT stash during retry attempts - only on final failure
+- Use descriptive stash name: `raf-{project-number}-task-{task-number}-failed`
+- Log message telling user about the stash with the stash name
+- Example log: `Changes stashed as: raf-001-task-3-failed`
+
+## Implementation Steps
+1. Modify `src/core/git.ts` to add a `stashChanges(name: string)` function
+2. In `src/commands/do.ts`, after a task is marked as completely failed (no more retries):
+   - Check if there are uncommitted changes using existing `getChangedFiles()`
+   - If changes exist, call `stashChanges()` with formatted name
+   - Log the stash name to inform the user
+3. Format stash name as: `raf-{projectNumber}-task-{taskId}-failed`
+4. Use `git stash push -m "{stash-name}"` command
+
+## Acceptance Criteria
+- [ ] Changes are stashed when task fails completely (after all retries)
+- [ ] Changes are NOT stashed during retry attempts
+- [ ] Stash name follows format: `raf-001-task-3-failed`
+- [ ] User sees log message with stash name
+- [ ] No stash created if no uncommitted changes exist
+- [ ] All existing tests pass
+
+## Notes
+- Only stash if there are actual changes (check with `git status`)
+- The stash preserves both staged and unstaged changes
+- User can recover with `git stash list` and `git stash pop stash@{n}`
+- Consider adding the stash name to the task failure outcome for reference

package/RAF/001-raf-task-improvements/plans/004-add-project-name-to-commits.md

@@ -0,0 +1,34 @@
+# Task: Add Project Name to Commit Messages
+
+## Objective
+Include the project name in commit messages using the format `RAF(project-name): task message`.
+
+## Context
+Currently, commits made by RAF don't include project context in the message. This makes it harder to identify which project a commit belongs to when looking at git history. Adding a consistent prefix improves traceability.
+
+## Requirements
+- Commit message format: `RAF(project-name): original message`
+- Example: `RAF(my-project): Fix login validation bug`
+- Apply to all automatic commits made during task execution
+- Project name should be extracted from project folder name (without the number prefix)
+
+## Implementation Steps
+1. Modify `src/core/git.ts` `commitTaskChanges()` function
+2. Add a parameter for project name or pass it through the commit context
+3. Extract project name from folder: `001-my-project` -> `my-project`
+4. Format commit message as: `RAF({projectName}): {originalMessage}`
+5. Update `src/commands/do.ts` to pass project name when calling commit
+6. Update execution prompt in `src/prompts/execution.ts` to inform Claude about the new commit format
+
+## Acceptance Criteria
+- [ ] All automatic commits use format `RAF(project-name): message`
+- [ ] Project name is correctly extracted from folder name (strips number prefix)
+- [ ] Works with various project folder name formats
+- [ ] Existing commit functionality still works
+- [ ] All tests pass
+
+## Notes
+- The project folder format is `NNN-project-name` where NNN is a 3-digit number
+- Use regex like `/^\d+-(.+)$/` to extract the name portion
+- Make sure special characters in project names don't break the commit
+- The original commit message from Claude should be preserved after the prefix

package/RAF/001-raf-task-improvements/plans/005-add-running-time-display.md

@@ -0,0 +1,39 @@
+# Task: Add Running Time Display During Task Execution
+
+## Objective
+Show a live timer in the status line during task execution, displaying elapsed time like `⏱ 2m 34s`.
+
+## Context
+Currently, there's no visibility into how long a task has been running during the "do" phase. A live timer helps users understand task progress and identify potential hangs or long-running operations.
+
+## Requirements
+- Show continuously updating timer during task execution
+- Format: `⏱ 2m 34s` or similar human-readable format
+- Update frequency: every second
+- Display in status line (not as log spam)
+- Timer starts when task begins, stops when task completes or fails
+
+## Implementation Steps
+1. Create a timer utility in `src/utils/timer.ts` or similar
+2. Implement a status line that can be updated in-place (using `\r` or cursor control)
+3. In `src/commands/do.ts`, start timer when task execution begins
+4. Update the display every second with current elapsed time
+5. Stop and clear timer when task completes
+6. Format time as: `Xm Ys` for minutes/seconds, `Xh Ym` for hours/minutes
+7. Ensure timer doesn't interfere with Claude's output streaming
+
+## Acceptance Criteria
+- [ ] Live timer shown during task execution
+- [ ] Timer updates every second
+- [ ] Format is human-readable (e.g., `⏱ 2m 34s`)
+- [ ] Timer stops when task completes or fails
+- [ ] Timer doesn't interfere with other output
+- [ ] Works correctly with verbose mode
+- [ ] All tests pass
+
+## Notes
+- Use `setInterval` for timer updates, clear on completion
+- Consider using `process.stdout.write('\r...')` for in-place updates
+- May need to handle terminal width for proper display
+- Timer should be cleared/hidden before final status output
+- Consider storing elapsed time in task outcome for summary

package/RAF/001-raf-task-improvements/plans/006-add-task-name-to-logs.md

@@ -0,0 +1,37 @@
+# Task: Add Task Name to Execution Logs
+
+## Objective
+Display task number and name in log messages during project execution, formatted as `[Task 2/5: fix-login-bug]`.
+
+## Context
+During task execution, log messages don't indicate which task they belong to. This makes it difficult to track progress and debug issues when multiple tasks run. Adding task context to logs improves clarity.
+
+## Requirements
+- Log format: `[Task X/Y: task-name]` where X is current task, Y is total tasks
+- Example: `[Task 2/5: fix-login-bug]`
+- Apply to relevant log messages during task execution
+- Task name extracted from plan file name (e.g., `002-fix-login-bug.md` -> `fix-login-bug`)
+
+## Implementation Steps
+1. Modify `src/utils/logger.ts` to support a context prefix
+2. Add method like `logger.setContext(prefix: string)` or `logger.withContext(prefix)`
+3. In `src/commands/do.ts`, when starting a task:
+   - Extract task name from plan file (strip number prefix and extension)
+   - Format prefix as `[Task {current}/{total}: {taskName}]`
+   - Set logger context with this prefix
+4. Clear context when task completes
+5. Update relevant log calls to include the context
+
+## Acceptance Criteria
+- [ ] Task context shown in logs during execution: `[Task 2/5: fix-login-bug]`
+- [ ] Task number shows current/total (e.g., 2/5)
+- [ ] Task name correctly extracted from plan filename
+- [ ] Context cleared between tasks
+- [ ] Doesn't affect non-task logging
+- [ ] All tests pass
+
+## Notes
+- Plan files are named like `001-task-name.md`, `002-another-task.md`
+- Total task count comes from `state.tasks.length`
+- Current task number is `taskIndex + 1` (1-indexed for display)
+- Consider whether all logs or only specific ones should show context

package/RAF/001-raf-task-improvements/plans/009-remove-editor-placeholder-text.md

@@ -0,0 +1,34 @@
+# Task: Remove Editor Placeholder Text
+
+## Objective
+Open the editor with an empty file instead of placeholder text "Describe your project here..." during the plan phase.
+
+## Context
+Currently, when running `raf plan`, the editor opens with placeholder text that the user must delete before writing their project description. Opening with an empty file provides a cleaner experience.
+
+## Requirements
+- Editor opens with completely empty file (no placeholder text)
+- No template, hints, or comments in the file
+- User starts with a blank canvas
+- Works with all supported editors ($EDITOR)
+
+## Implementation Steps
+1. Locate the editor template in `src/core/editor.ts` or `src/commands/plan.ts`
+2. Find where the initial content/template is defined
+3. Remove or comment out the placeholder text
+4. Ensure the file is created but with empty content (`''`)
+5. Test with different editors to ensure empty file works
+
+## Acceptance Criteria
+- [ ] `raf plan` opens editor with empty file
+- [ ] No placeholder text present
+- [ ] No template or comments
+- [ ] Works with vim, nano, code, and other common editors
+- [ ] All tests pass
+
+## Notes
+- This is a simple change - just remove the template content
+- The placeholder text is likely defined as a string constant somewhere
+- Make sure the file is still created (just empty)
+- User experience is cleaner without needing to delete boilerplate
+- If there were useful hints, they could be shown in terminal before editor opens instead

package/RAF/002-raf-task-improvements-execution/decisions/DECISIONS.md

@@ -0,0 +1,13 @@
+# Project Decisions
+
+## For Task 1 (Show Model at Task Start): The original plan mentions detecting the model from Claude CLI. Do you have a specific method in mind?
+Task is already in progress - implementation exists in uncommitted source files. Uses `getClaudeModel()` which reads from `~/.claude/settings.json`.
+
+## Task 1 (Show Model) is already implemented. Should this project just commit it?
+Include task to commit the existing uncommitted changes for Task 007.
+
+## For Task 2 (Move State/Logs to .raf/): Should we migrate existing state?
+Skip this task entirely - user wants to get rid of .raf completely in the future.
+
+## What should we do with the 008-move-state-logs-to-project.md plan file?
+Delete it - the task won't be implemented.

package/RAF/002-raf-task-improvements-execution/input.md

@@ -0,0 +1,3 @@
+move RAF/001-raf-task-improvements/plans/007-show-model-at-task-start.md and RAF/001-raf-task-improvements/plans/008-move-state-logs-to-project.md
+into new raf project for execution
+

package/RAF/002-raf-task-improvements-execution/outcomes/001-commit-show-model-at-task-start.md

@@ -0,0 +1,17 @@
+# Task 001 - Completed
+
+## Summary
+All acceptance criteria have been met:
+1. ✅ All tests pass (177 tests passing)
+2. ✅ Changes committed with descriptive message
+3. ✅ Version updated in package.json (0.2.5 → 0.2.6)
+4. ✅ Plan file moved to outcomes folder
+5. ✅ Outcome file documents the implementation
+The task has been completed successfully.
+<promise>COMPLETE</promise>
+
+## Details
+- Attempts: 1
+- Elapsed time: 1m 15s
+- Completed at: 2026-01-30T17:31:19.872Z
+- Commit: 904a1de

package/RAF/002-raf-task-improvements-execution/outcomes/002-delete-skipped-plan.md

@@ -0,0 +1,23 @@
+# Task 002 - Completed
+
+## Summary
+Let me verify all acceptance criteria are met:
+1. ✅ Plan file 008-move-state-logs-to-project.md is deleted
+2. ✅ No orphaned references to this plan exist (only documentation references in DECISIONS.md, input.md, and this plan which are expected)
+3. ✅ Tests pass (177 tests passing)
+4. ✅ Changes committed
+# Task 002 - Completed
+## Summary
+All acceptance criteria have been met:
+1. ✅ Plan file `008-move-state-logs-to-project.md` is deleted
+2. ✅ No orphaned references exist (only documentation references)
+3. ✅ Updated SUMMARY.md to mark Task 008 as skipped
+4. ✅ All 177 tests pass
+5. ✅ Changes committed with descriptive message
+<promise>COMPLETE</promise>
+
+## Details
+- Attempts: 1
+- Elapsed time: 1m 5s
+- Completed at: 2026-01-30T17:32:25.642Z
+

package/RAF/002-raf-task-improvements-execution/outcomes/SUMMARY.md

@@ -0,0 +1,32 @@
+# Project Summary: raf-task-improvements-execution
+
+**Status:** executing
+**Created:** 2026-01-30T17:21:17.303Z
+**Updated:** 2026-01-30T17:32:25.642Z
+
+## Statistics
+
+- Completed: 2
+- Failed: 0
+- Skipped: 0
+- Pending: 0
+
+## Tasks
+
+### [x] Task 001
+
+- **Plan:** plans/001-commit-show-model-at-task-start.md
+- **Status:** completed
+- **Attempts:** 1
+- **Started:** 2026-01-30T17:30:04.073Z
+- **Completed:** 2026-01-30T17:31:19.872Z
+- **Commit:** 904a1de
+
+### [x] Task 002
+
+- **Plan:** plans/002-delete-skipped-plan.md
+- **Status:** completed
+- **Attempts:** 1
+- **Started:** 2026-01-30T17:31:19.894Z
+- **Completed:** 2026-01-30T17:32:25.642Z
+

package/RAF/002-raf-task-improvements-execution/plans/001-commit-show-model-at-task-start.md

@@ -0,0 +1,37 @@
+# Task: Commit Show Model at Task Start
+
+## Objective
+Commit the existing implementation that logs the Claude model name when task execution begins.
+
+## Context
+Task 007 from RAF/001-raf-task-improvements has already been implemented. The code changes exist in uncommitted files and need to be committed to complete the task.
+
+## Requirements
+- Commit the existing implementation in `src/commands/do.ts` and `src/utils/config.ts`
+- Include the test file `tests/unit/config.test.ts`
+- Verify tests pass before committing
+- Update version in package.json
+- Move the original plan file to outcomes folder
+
+## Implementation Steps
+1. Run tests to verify the implementation works: `npm test`
+2. Review the uncommitted changes:
+   - `src/utils/config.ts`: Added `getClaudeModel()` function that reads model from `~/.claude/settings.json`
+   - `src/commands/do.ts`: Added logging of model name at task start
+   - `tests/unit/config.test.ts`: Unit tests for `getClaudeModel()`
+3. Update version in package.json (minor bump)
+4. Stage and commit the changes with appropriate commit message
+5. Move `RAF/001-raf-task-improvements/plans/007-show-model-at-task-start.md` to `RAF/001-raf-task-improvements/outcomes/007-show-model-at-task-start.md`
+6. Create outcome summary documenting what was implemented
+
+## Acceptance Criteria
+- [ ] All tests pass
+- [ ] Changes committed with descriptive message
+- [ ] Version updated in package.json
+- [ ] Plan file moved to outcomes folder
+- [ ] Outcome file documents the implementation
+
+## Notes
+- The implementation reads model from Claude CLI settings at `~/.claude/settings.json`
+- Returns null gracefully if settings file doesn't exist or model not specified
+- Model is logged once at task start: `Using model: {model-name}`

package/RAF/002-raf-task-improvements-execution/plans/002-delete-skipped-plan.md

@@ -0,0 +1,23 @@
+# Task: Delete Skipped Plan File
+
+## Objective
+Remove the 008-move-state-logs-to-project.md plan file since the task will not be implemented.
+
+## Context
+The user decided to skip the "Move State and Logs to .raf/" task entirely. They plan to eliminate .raf usage completely in the future, so this plan is no longer relevant.
+
+## Requirements
+- Delete the plan file from RAF/001-raf-task-improvements/plans/
+- Do not create any outcome file for this skipped task
+
+## Implementation Steps
+1. Delete `RAF/001-raf-task-improvements/plans/008-move-state-logs-to-project.md`
+2. Verify the file has been removed
+
+## Acceptance Criteria
+- [ ] Plan file 008-move-state-logs-to-project.md is deleted
+- [ ] No orphaned references to this plan exist
+
+## Notes
+- This task was intentionally skipped per user decision, not abandoned
+- The decision is documented in this project's DECISIONS.md

package/RAF/003-multi-project-execution/decisions/DECISIONS.md

@@ -0,0 +1,68 @@
+# Project Decisions
+
+## Task 1: Multi-Project Execution
+
+### For multi-project execution, should projects run sequentially or in parallel?
+**Answer:** Sequential - Run projects one after another, safer for git commits
+
+### When running multiple projects, if one fails should we continue with remaining projects?
+**Answer:** Continue - Continue with remaining projects, report failures at end
+
+### For number-based selection (e.g., 'raf do 003 004'), should we support ranges like '003-005'?
+**Answer:** No, individual only - Only support listing individual numbers
+
+## Task 2: Auto-Commit After Planning
+
+### For auto-commit after planning, what commit message format should be used?
+**Answer:** RAF(project-name): Plan complete - Consistent with current task commit format
+
+## Task 3: Commit Outcomes with Changes
+
+### Should the outcome be committed in the SAME commit as task changes?
+**Answer:** Same commit - Outcome file included in task's commit
+
+## Task 4: Re-Run Failed Tasks / State Derivation
+
+### How should we determine if a task is 'completed'?
+**Answer:** Outcome file exists - Task is complete if outcomes/NNN-task.md exists with success marker
+
+### Should we completely remove the .raf folder and state.json?
+**Answer:** Yes, remove completely - Derive everything from RAF/project folders
+
+### When re-running 'raf do project-name' on a project with failures, what should happen?
+**Answer:** Resume from failed/pending - Skip completed tasks, retry failed and pending tasks
+
+## Task 5: Simplify Git Logic
+
+### Should Claude be instructed to commit, or should RAF commit after task completion?
+**Answer:** Claude commits (instructed) - Add instructions for Claude to commit after making changes
+
+### Should the 'smart commit' logic (baseline tracking, filtering) be simplified or removed?
+**Answer:** Remove, commit all - Remove smart filtering, commit all changes after task
+
+### When Claude is instructed to commit, what commit message format should Claude use?
+**Answer:** [project] task description (no numbers) - Custom format specified by user
+
+## Task 6: Per-Task Timeout
+
+### Is the timeout issue that multiple retries extend beyond 60 min total?
+**Answer:** Just verify/document - Verify current implementation and add tests
+
+## Additional Decisions
+
+### Where should execution logs be stored?
+**Answer:** No logs, console only - Don't persist logs, show in terminal only
+
+### What marker should indicate success vs failure in outcome files?
+**Answer:** Content marker at top - "## Status: SUCCESS" or "## Status: FAILED" in file
+
+### How should 'raf status' work without state.json?
+**Answer:** Scan RAF folder, derive status - List projects by scanning RAF/, derive status from files
+
+### Should RAF support running a project that's already fully completed?
+**Answer:** Allow with --force flag - Require --force flag to re-run completed project
+
+## Task 8: Move Decisions File
+
+### Should decisions be stored in a folder or as a single file?
+**Answer:** Single file at project root - Move from `decisions/DECISIONS.md` to `decisions.md`

package/RAF/003-multi-project-execution/input.md

@@ -0,0 +1,6 @@
+- [ ] should be able to pass multiple project names or number to "raf do", like "raf do projectName otherProjectName" or "raf do 003 004 005"
+- [ ] after plan step comlite - commit the changes in RAF (input, decisions, plans) for current project (commit project new project folder basically)
+- [ ] outcomes file for project should be commited together with changes
+- [ ] let user re-run failed tasks by doing "raf do project-name" command again, currently it's outputs statistic. if there are failied task or unfinished - do them. use case - takens limit reached or api unavalibility etc. we are relying on state too much, rely on .md documents instead (use claude code with sonnet model internally to derive state). i think it should be part of getting rid of ./raf technical folder and state as a concept, all state should be derived from folder structure, claude code should be used internally when it's hard to derive info from convention
+- [ ]  git logic (code) should be simplified, running task should commit the changes it made, add certain instructions for that
+- [ ] ensure that timeout (60 min) is per task

package/RAF/003-multi-project-execution/outcomes/001-remove-state-json.md

@@ -0,0 +1,52 @@
+# Task 001 - Completed
+
+## Summary
+The task has been completed successfully. Let me provide a summary:
+## Task 001 Complete
+### Summary
+Removed the `.raf` folder and `state.json` entirely, replacing them with a state derivation approach that derives all project state from the folder structure.
+### Changes Made
+1. **Created `src/core/state-derivation.ts`** - New module that:
+   - Parses outcome files for `## Status: SUCCESS` or `## Status: FAILED` markers
+   - Derives project state from `plans/` and `outcomes/` directories
+   - Provides functions: `deriveProjectState`, `getNextPendingTask`, `getDerivedStats`, `isProjectComplete`, `hasProjectFailed`
+2. **Updated `src/commands/do.ts`**:
+   - Uses state derivation instead of StateManager
+   - Outcome files now include status markers at the top
+   - Re-derives state after each task completion
+3. **Updated `src/commands/plan.ts`**:
+   - Removed StateManager dependency
+   - Simply checks for plan files directly
+4. **Updated `src/commands/status.ts`**:
+   - Uses state derivation for project status display
+5. **Updated `src/core/project-manager.ts`**:
+   - `createProject` now returns just the path (not StateManager)
+   - `saveSummary` accepts derived state instead of StateManager
+   - `listProjects` includes derived task counts
+   - Logs saved to project folder (`logs/`) instead of `.raf/logs/`
+6. **Updated `src/core/shutdown-handler.ts`**:
+   - Removed StateManager registration and save on shutdown
+7. **Updated `src/utils/paths.ts`**:
+   - Removed all `.raf` related utilities (`RAF_RUNTIME_DIR`, `getRafRuntimeDir`, `ensureRafRuntimeDir`, `getRuntimeStatePath`, `getRuntimeLogsDir`, `ensureRuntimeLogsDir`, `getStatePath`)
+8. **Removed**:
+   - `src/core/state-manager.ts`
+   - `src/types/state.ts`
+   - `tests/unit/state-manager.test.ts`
+9. **Added/Updated Tests**:
+   - Created `tests/unit/state-derivation.test.ts` with 20 test cases
+   - Updated `tests/unit/project-manager.test.ts` to work with new API
+### Acceptance Criteria Met
+- ✅ `.raf` folder is not created or used by any command
+- ✅ `state.json` is not created or read
+- ✅ Project state can be derived from `RAF/project/` folder structure
+- ✅ Outcome files contain `## Status: SUCCESS` or `## Status: FAILED` marker
+- ✅ All existing tests updated or removed as appropriate
+- ✅ New tests for state derivation logic
+- ✅ Code compiles without errors (178 tests passing)
+<promise>COMPLETE</promise>
+
+## Details
+- Attempts: 1
+- Elapsed time: 7m 32s
+- Completed at: 2026-01-30T18:03:28.628Z
+

package/RAF/003-multi-project-execution/outcomes/002-update-raf-status.md

@@ -0,0 +1,50 @@
+# Task 002 - Completed
+
+## Summary
+## Task 002 Complete
+### Summary
+Updated the `raf status` command to work without `state.json`, deriving all status information from the RAF folder structure and outcome files.
+### Changes Made
+1. **Updated `src/core/state-derivation.ts`**:
+   - Added `DerivedProjectStatus` type with values: `planning`, `ready`, `executing`, `completed`, `failed`
+   - Added `DiscoveredProject` interface for discovered projects
+   - Added `discoverProjects(rafDir)` function to scan RAF directory and return all projects matching `NNN-project-name` pattern
+   - Added `deriveProjectStatus(projectPath, tasks)` function to derive project status from tasks:
+     - `planning`: no plans folder or empty plans
+     - `ready`: has plan files but all tasks are pending
+     - `executing`: some tasks completed, others pending
+     - `completed`: all tasks completed successfully
+     - `failed`: any task has failed
+   - Updated `DerivedProjectState` interface to include `status` field
+   - Updated `deriveProjectState()` to return status along with tasks
+2. **Updated `src/commands/status.ts`**:
+   - Replaced `listProjects` import with `discoverProjects` from state-derivation
+   - Simplified project status determination using `state.status` instead of inline computation
+   - Updated `getProjectStatusBadge()` to support all project statuses:
+     - `[P]` for planning
+     - `[R]` for ready
+     - `[~]` for executing
+     - `[x]` for completed
+     - `[!]` for failed
+   - Removed unused imports (`isProjectComplete`, `hasProjectFailed`)
+3. **Updated `tests/unit/state-derivation.test.ts`**:
+   - Added 4 tests for `discoverProjects()` function
+   - Added 6 tests for `deriveProjectStatus()` function
+   - Added 4 new tests for `deriveProjectState()` to verify status derivation
+   - Updated all existing tests to include `status` field in test state objects
+### Acceptance Criteria Met
+- ✅ `raf status` lists all projects from RAF folder (uses `discoverProjects()`)
+- ✅ `raf status <project>` shows detailed task status derived from files
+- ✅ Status badges display correctly: `[ ]` pending, `[x]` completed, `[!]` failed
+- ✅ `--json` flag outputs valid JSON with derived state including project status
+- ✅ No dependency on StateManager or state.json
+- ✅ Tests cover status derivation scenarios (34 tests passing)
+### Test Results
+All 191 tests pass (including 34 tests for state-derivation).
+<promise>COMPLETE</promise>
+
+## Details
+- Attempts: 1
+- Elapsed time: 3m 50s
+- Completed at: 2026-01-30T18:07:19.042Z
+

package/RAF/003-multi-project-execution/outcomes/003-simplify-git-logic.md

@@ -0,0 +1,35 @@
+# Task 003 - Completed
+
+## Summary
+## Task 003 Complete
+### Summary
+Simplified git operations by removing smart commit logic (baseline tracking, delta filtering) from RAF and instructing Claude to commit its own changes during task execution.
+### Changes Made
+1. **Updated `src/prompts/execution.ts`**:
+   - Added `outcomeFilePath` optional parameter
+   - Updated commit instructions to use format `[project-name] brief description` (no task numbers)
+   - Added note about outcome file being written by RAF after Claude's commit
+2. **Simplified `src/core/git.ts`**:
+   - Removed: `commitChanges`, `getLastCommitHash`, `getTaskChangedFiles`, `commitSpecificFiles`, `formatCommitMessage`, `commitTaskChanges`, `FileChange` interface
+   - Kept: `isGitRepo`, `getGitStatus`, `hasUncommittedChanges`, `parseGitStatus`, `getChangedFiles`, `stashChanges`
+3. **Updated `src/commands/do.ts`**:
+   - Removed baseline tracking and commit logic
+   - Kept stash functionality for failure cases
+4. **Updated `tests/unit/git.test.ts`**:
+   - Removed tests for removed functions
+   - Kept tests for `parseGitStatus`
+### Acceptance Criteria Met
+- ✅ Execution prompt includes commit instructions for Claude
+- ✅ Commit message format is `[project-name] task description` (no task numbers)
+- ✅ Smart commit logic (baseline tracking) is removed
+- ✅ RAF does not commit after task completion
+- ✅ Outcome file is noted in prompt (written by RAF after Claude's work)
+- ✅ Stash functionality still works on task failure
+- ✅ All tests pass (178 tests)
+<promise>COMPLETE</promise>
+
+## Details
+- Attempts: 1
+- Elapsed time: 3m 38s
+- Completed at: 2026-01-30T18:10:57.209Z
+

package/RAF/003-multi-project-execution/outcomes/004-auto-commit-planning.md

@@ -0,0 +1,43 @@
+# Task 004 - Completed
+
+## Summary
+The remaining change is to the outcome file from task 003, which is not part of this task.
+## Task 004 Complete
+### Summary
+Implemented auto-commit functionality after the `raf plan` command completes successfully, committing only the project folder files.
+### Changes Made
+1. **Added `commitProjectFolder` function to `src/core/git.ts`**:
+   - Stages only files within the specified project folder
+   - Uses commit message format: `RAF(project-name): Plan complete`
+   - Returns `CommitResult` with success status, message, or error
+   - Handles edge case of no changes to commit
+2. **Updated `src/commands/plan.ts`**:
+   - Calls `commitProjectFolder` after plan files are created successfully
+   - Logs success or failure to user
+   - Does not fail planning if commit fails (graceful degradation)
+3. **Added `tests/unit/git-commit-project.test.ts`**:
+   - 7 test cases covering all commit scenarios:
+     - Not in git repo
+     - Successful commit with correct message format
+     - Only project files staged
+     - No changes to commit
+     - Git add failure
+     - Git commit failure
+     - Path escaping
+### Acceptance Criteria Met
+- ✅ After successful `raf plan`, project folder is committed
+- ✅ Commit message is `RAF(project-name): Plan complete`
+- ✅ Only files in `RAF/NNN-project-name/` are committed
+- ✅ Unrelated changes in repo are not committed
+- ✅ Planning doesn't fail if commit fails
+- ✅ User is informed of commit success/failure
+- ✅ Tests cover commit scenarios (7 tests)
+### Test Results
+All 185 tests pass.
+<promise>COMPLETE</promise>
+
+## Details
+- Attempts: 1
+- Elapsed time: 2m 21s
+- Completed at: 2026-01-30T18:13:18.833Z
+

package/RAF/003-multi-project-execution/outcomes/005-rerun-failed-tasks.md

@@ -0,0 +1,43 @@
+# Task 005 - Completed
+
+## Summary
+The task is complete. Let me provide a summary:
+## Task 005 Complete
+### Summary
+Implemented support for re-running failed and pending tasks with the `raf do` command. Users can now resume project execution from failed or pending tasks, and use `--force` to re-run all tasks.
+### Changes Made
+1. **Updated `src/types/config.ts`**:
+   - Added `force?: boolean` to `DoCommandOptions` interface
+2. **Updated `src/commands/do.ts`**:
+   - Added `-f, --force` flag to re-run all tasks regardless of status
+   - Changed from `getNextPendingTask` to `getNextExecutableTask` for task selection
+   - Implemented force mode that tracks tasks completed in the current session
+   - Added console output indicating task status:
+     - "Retrying task X (previously failed)..." for failed tasks
+     - "Re-running task X (force mode)..." for completed tasks with --force
+     - "Executing task X..." for pending tasks
+   - Updated "project complete" message to suggest `--force` flag
+   - Track completed tasks in session for force mode iteration
+3. **Created `tests/unit/do-rerun.test.ts`**:
+   - 11 tests covering all rerun scenarios:
+     - Skip completed tasks
+     - Retry failed tasks
+     - Run pending tasks
+     - Handle malformed outcome files
+     - All complete detection
+     - Force mode task selection
+     - Task status logging
+### Acceptance Criteria Met
+- ✅ `raf do project-name` skips tasks with SUCCESS outcome
+- ✅ `raf do project-name` retries tasks with FAILED outcome
+- ✅ `raf do project-name` runs tasks without outcome files
+- ✅ `--force` flag runs all tasks regardless of status
+- ✅ Clear console output indicates skip/retry/run status
+- ✅ All tests pass (196 tests)
+<promise>COMPLETE</promise>
+
+## Details
+- Attempts: 1
+- Elapsed time: 3m 1s
+- Completed at: 2026-01-30T18:16:20.525Z
+