document-hoarder

# Document Hoarder

Fetch Confluence pages and organize them into structured local documentation.

## Prerequisites

- atlassian skill must be available for Confluence API access
- Environment variables configured: ATLASSIAN_USER_EMAIL, ATLASSIAN_API_TOKEN, JIRA_URL

## Workflow

### Step 1: Identify Target

Accept one of:
- Confluence page URL (extract page ID from path `/pages/{page_id}/`)
- Page ID directly
- Local file path to update (read frontmatter for `page_id`)
- Document name to search

### Step 2: Fetch Page

```bash
uv run --no-project --with requests python /home/drew/.claude/skills/atlassian/scripts/confluence_api.py read {page_id} -o {output_path}
```

For multiple pages, spawn sub-agents in parallel:
```
Task(subagent_type="general-purpose", run_in_background=true, model="haiku")
```

Each sub-agent should:
1. Fetch the page using confluence_api.py
2. Read the saved file
3. Extract any Confluence links (pattern: `atlassian.net/wiki/spaces/.*/pages/(\d+)`)
4. Return: brief summary + list of found links

### Step 3: Handle Linked Pages

After sub-agents complete:
1. Collect all discovered Confluence links
2. Filter out already-fetched pages
3. Ask user: "Found N linked pages. Fetch them all?"
   - Yes: repeat Step 2 for new pages
   - No: proceed to Step 4
   - Selective: let user choose which to fetch

### Step 4: Organize Structure

After all pages fetched:
1. List all saved markdown files
2. Analyze content to determine logical grouping:
   - `core/` - operational documents
   - `training/` - tutorials, guides
   - `reference/` - technical references
   - `meetings/` - meeting notes, schedules
3. Create subdirectories and move files
4. Generate `index.md` with:
   - Folder structure overview
   - Links to all documents with descriptions
   - Recommended reading order if applicable

## Update Workflow

When user requests update to existing docs:

1. Identify target:
   - Specific file path: read its `page_id` from frontmatter
   - Folder: scan all .m