---
name: process-notes
description: Process CEO notes from docs/CEO_NOTES.md into GitHub issues, design decisions, blog topics, and sprint items. Reads raw thoughts and turns them into actionable work.
allowed-tools: Bash, Read, Write, Edit, Grep, Glob, Task
argument-hint: [--dry-run]
---

# Process Notes — Turn Thoughts Into Action

## Overview

Reads `docs/CEO_NOTES.md` Inbox section and processes each note into one or more of:
- **GitHub Issue** — feature requests, bugs, technical tasks
- **Design Decision** — write to `intel.decisions` in Supabase
- **Blog Topic** — queue for the `/writer` agent
- **Sprint Item** — tag for next sprint planning
- **Question** — needs more context, ask the CEO

## Workflow

### Step 0: Harvest Scratchpad Ideas (Auto)

Before reading CEO_NOTES.md, check `.claude/SCRATCH.md` for pending ideas:

```
Read .claude/SCRATCH.md
Find the "### Ideas → CEO Notes" section
Extract any non-empty bullet items (lines starting with "- " that aren't just "- ")
If items found:
  1. Append each item as a new bullet under "## Inbox" in docs/CEO_NOTES.md
  2. Clear the "### Ideas → CEO Notes" section back to a single blank placeholder:
       ### Ideas → CEO Notes
       -
  3. Print: "Harvested N idea(s) from SCRATCH.md → CEO_NOTES.md Inbox"
If no items found: skip silently
```

This keeps the scratchpad ephemeral while ensuring ideas aren't lost. Items are moved, not copied.

### Step 1: Read the Notes

```
Read docs/CEO_NOTES.md
Extract all items from the ## Inbox section (between --- markers)
Skip empty lines and HTML comments
```

### Step 2: Categorize Each Note

For each note, determine category using this logic:

| Signal | Category | Action |
|--------|----------|--------|
| "bug", "broken", "wrong", "fix" | Bug | Create `[FIX]` GitHub issue, label `type:fix` |
| "idea", "feature", "should", "need", "want" | Feature | Create `[FEAT]` GitHub issue, label `type:feature` |
| "decide", "should we", "option", "vs" | Decision | Write to `intel.decisions` in Supabase as OPEN |
| "blog", "write about", "article", "post" | Blog | Add to `.claude/content/blog-queue.md` |
| "priority", "important", "urgent", "p0" | Priority Shift | Flag for `/pm` review |
| "question", "?", "ask", "wonder" | Question | Present to user via AskUserQuestion |

### Step 3: Execute Actions

**For GitHub Issues:**
```bash
gh issue create --title "[TYPE] Description" \
  --body "Source: CEO Notes ($(date +%Y-%m-%d))\n\n$NOTE_TEXT" \
  --label "type:feature,source:ceo-notes"
```

**For Blog Topics:**
- Append to `.claude/content/blog-queue.md`
- Include: topic, angle, key points, target platform (LinkedIn/Medium/blog)

**For Design Decisions:**
- Find next decision number in `intel.decisions` in Supabase
- Add as OPEN with the note as context
- Present for CEO confirmation

### Step 4: Move to Processed

After processing each note:
1. Remove from Inbox section
2. Add to Processed section with timestamp and links:
   ```
   - [2026-02-25] "Original note text" → #500 (issue), blog queue
   ```

### Step 5: Summary

Present a summary:
```
Processed 7 notes:
  3 → GitHub issues (#500, #501, #502)
  2 → Blog queue (Medium articles)
  1 → Design decision (OPEN #265)
  1 → Question (asked above)
```

## Dry Run Mode

With `--dry-run`, show what WOULD be created without actually creating anything.
Useful for reviewing categorization before committing.

## Safety

- Never deletes notes — moves them to Processed
- Never creates duplicate issues (checks existing titles first)
- Always shows what it will do before doing it
- Commits the updated CEO_NOTES.md after processing