# Save Project Plan Command

You are helping the user create a comprehensive project plan recap that will be saved to `.claude/implementation/NEXT_SESSION.md`. This is a **single file** that gets completely overwritten each session to provide the next AI agent with maximum context.

## Your Task

Automatically analyze the current project state and generate a complete project plan document for the next coding session.

## Analysis Steps

### 1. Gather Git Context
Run these commands to understand recent work:
```bash
# Get last 10 commits with detailed info
git log -10 --pretty=format:"%h - %s (%ar)" --no-merges

# Get current branch and status
git status --short

# Get files changed in last 5 commits
git diff --name-status HEAD~5..HEAD

# Get current branch name
git branch --show-current
```

### 2. Read Current Plan
Read `.claude/implementation/NEXT_SESSION.md` to understand:
- What was planned for this session
- What completion % we started at
- What tasks were on the list

### 3. Read Implementation Index
Read `.claude/implementation/00-index.md` to understand:
- Overall project status
- Which phase/week we're in
- What's completed vs. pending

### 4. Analyze File Changes
For the most recently modified files (from git diff), determine:
- What features were implemented
- What tests were added
- What architectural changes were made

## Document Structure

Generate a NEW `NEXT_SESSION.md` file with this exact structure:

```markdown
# Next Session Plan - [Phase/Week Name]

**Current Status**: [Phase] - [X]% Complete
**Last Commit**: `[hash]` - "[commit message]"
**Date**: [YYYY-MM-DD]
**Remaining Work**: [X]% ([N] tasks)

---

## Quick Start for Next AI Agent

### 🎯 Where to Begin
1. Read this entire document first
2. Review files in "Files to Review Before Starting" section
3. Start with Task 1 in "Tasks for Next Session"
4. Run test commands after each task

### 📍 Current Context
[2-3 sentences describing exactly where we are in the project and what the immediate focus is]

---

## What We Just Completed ✅

[Analyze git commits and file changes to document what was done this session]

### 1. [Feature/Component Name]
   - `path/to/file.py` - [What changed]
   - [Specific implementation details]
   - [N] tests passing

### 2. [Next Feature]
   - [Details...]

[Continue for all major changes]

---

## Key Architecture Decisions Made

[Extract any important architectural choices, patterns adopted, or technical decisions]

1. **[Decision Name]**: [Rationale and implications]
2. **[Decision Name]**: [Rationale and implications]

---

## Blockers Encountered 🚧

[List any issues that blocked progress or need attention]

- **[Blocker Title]**: [Description, current status, what needs to be done]

[If none, write "None - development proceeded smoothly"]

---

## Outstanding Questions ❓

[List unresolved questions or decisions that need to be made]

1. **[Question]**: [Context and why it matters]
2. **[Question]**: [Context and implications]

[If none, write "None at this time"]

---

## Tasks for Next Session

[Generate specific, actionable tasks with time estimates]

### Task 1: [Task Name] ([Time estimate])

**File(s)**: `path/to/file.py`

**Goal**: [Clear description of what needs to be done]

**Changes**:
1. [Specific change with code example if helpful]
2. [Next specific change]

**Files to Update**:
- `path/to/file1.py` - [What to change]
- `path/to/file2.py` - [What to change]

**Test Command**:
```bash
[Exact command to verify this task]
```

**Acceptance Criteria**:
- [ ] [Specific criterion]
- [ ] [Specific criterion]

---

### Task 2: [Task Name] ([Time estimate])

[Same structure as Task 1]

---

[Continue for all tasks - aim for 3-7 tasks]

---

## Files to Review Before Starting

[List critical files the next AI agent should read with specific line ranges if relevant]

1. `path/to/file.py:45-120` - [Why this is important]
2. `path/to/file2.py` - [Context about this file]
3. `.claude/implementation/[relevant-doc].md` - [What this contains]

---

## Verification Steps

After completing all tasks:

1. **Run all tests**:
   ```bash
   [Specific test commands]
   ```

2. **Manual testing** (if applicable):
   ```bash
   [Terminal client commands or manual steps]
   ```

3. **Database verification** (if applicable):
   ```sql
   [Specific queries to verify state]
   ```

4. **Commit changes**:
   ```bash
   git add [files]
   git commit -m "CLAUDE: [commit message]"
   ```

---

## Success Criteria

[Phase/Week] will be **[X]% complete** when:

- [ ] [Specific criterion with test count if applicable]
- [ ] [Specific criterion]
- [ ] [Specific criterion]
- [ ] Documentation updated
- [ ] Git commit created

---

## Quick Reference

**Current Test Count**: [N] tests ([breakdown by area])
**Last Test Run**: [Status] ([Date])
**Branch**: `[branch-name]`
**Python**: [version]
**Virtual Env**: `backend/venv/`

**Key Imports for Next Session**:
```python
[List the most important imports the next agent will need]
```

**Recent Commit History** (Last 10):
```
[Insert git log output from step 1]
```

---

## Context for AI Agent Resume

**If the next agent needs to understand the bigger picture**:
- Overall project: See `@prd-web-scorecard-1.1.md` and `@CLAUDE.md`
- Architecture: See `@.claude/implementation/00-index.md`
- Current phase details: See `@.claude/implementation/[current-phase].md`

**Critical files in current focus area**:
[List 5-10 most important files for the current work]

**What NOT to do**:
- [Any gotchas or common mistakes to avoid]
- [Patterns we've decided NOT to use and why]

---

**Estimated Time for Next Session**: [X] hours
**Priority**: [High/Medium/Low] ([Why])
**Blocking Other Work**: [Yes/No] ([Details if yes])
**Next Milestone After This**: [What comes next]
```

## Important Guidelines

1. **Be Specific**: Include actual file paths, line numbers, command examples
2. **Be Realistic**: Time estimates should account for testing and debugging
3. **Be Complete**: Future agent should not need to ask "what do I do next?"
4. **Include Context**: Explain WHY things were done, not just WHAT
5. **Flag Unknowns**: If something is unclear, document it in Outstanding Questions
6. **Test Commands**: Every task should have a specific test/verification command
7. **Link to Docs**: Reference other implementation docs with @ syntax when relevant

## After Generating

1. Write the complete document to `.claude/implementation/NEXT_SESSION.md`
2. Show the user a brief summary of:
   - Completion percentage
   - Number of commits analyzed
   - Number of tasks created for next session
   - Any critical blockers or questions
3. Ask if they want to add anything before finalizing

---

**Now begin the analysis and generate the project plan document.**