---
name: filing
description: Orchestrates file cleanup with mandatory processing — reads content and extracts actions BEFORE moving files. Prevents lost waiting-fors and buried actions. MANDATORY during weekly review filing phase. Triggers on 'where should this go', 'help me tidy', 'clean up downloads', 'triage inbox'. (user)
---

# Filing

Help with file organization: where things belong and keeping inboxes clear.

## When to Use

- "Where should this file go?"
- Weekly review (filing portion)
- Moving files between zones
- Deciding Projects vs Areas vs Resources

## When NOT to Use

- Pattern reflection (use todoist-gtd)
- Todoist organization (use todoist-gtd)
- Code/repo organization (different rules apply)

## Core Principle: Filing = Processing + Organizing

**You can't file what you haven't processed.**

Moving files without reading them isn't filing — it's relocating the problem. Proper filing means:

1. **Processing** — Read content, extract what matters (actions, waiting-fors, calendar items)
2. **Organizing** — Move to the right location with the right name

If you skip processing, actions get lost, waiting-fors vanish, and meeting notes become information graves.

**The test:** Before moving a file, can you answer: "What's in this and does anything need to happen?"

## Work Folder Structure

Location: `~/Library/CloudStorage/GoogleDrive-*/My Drive/Work`

To find exact path: `ls -d ~/Library/CloudStorage/GoogleDrive-*/My\ Drive/Work`

```
Work/
├── Projects/           # Active outcome folders
├── Areas/              # Ongoing responsibilities
├── Resources/          # Reference material (Larder, GTD Resources)
├── Archive/            # Completed/dormant
└── Meeting Notes/      # By year (2015-2025)
```

### Projects/
Active work with defined end states. Folders named: `{Area} - {Outcome statement}`

Examples:
- `Industry Influencing - Marketing Week Articles`
- `Product Development - Launched new dashboard...`

**When complete:** Move to Archive/

### Areas/
Ongoing responsibilities (no completion date):
- Budget and Finance
- Cross-Broadcaster Measurement
- Desired Outcomes Planning
- Industry Advisory Roles
- Industry Influencing
- Internal Stakeholders
- Managing Myself
- Processes and Systems
- Team Development

### Resources/
Reference material that supports work:
- **Larder/** - 100+ research PDFs
- **GTD Resources (Common)/** - GTD reference docs

### Archive/
Completed projects and dormant materials. Preserves history.

### Meeting Notes/
Organized by year:
```
Meeting Notes/
├── 2024/
└── 2025/
    └── 2025-12-11 Alex Maguire (Netflix) - EMEA measurement landscape.md
```

Naming: `{date} {Person/Topic} - {Brief description}.md`

## PARA Quick Reference

| Category | Definition | Completion |
|----------|------------|------------|
| **Projects** | Active work with end state | Finite - move to Archive when done |
| **Areas** | Ongoing responsibilities | Infinite - never "done" |
| **Resources** | Reference material | Supports work, isn't the work |
| **Archive** | Inactive items | Superseded or completed |

**Key distinction:** Areas ≠ Projects. "Team Development" is an Area (infinite). "Hire data scientist" is a Project (finite).

## How This Differs from Strict PARA

Strict PARA has 4 equal categories (Projects, Areas, Resources, Archive).

**This structure adds:**
- **Meeting Notes/** - Chronological organization by year (not in PARA)
- Areas are predefined (not emergent as in strict PARA)

**Core PARA principle remains:** Projects complete, Areas don't.

**Why the additions:** Meeting notes benefit from chronological access. Areas reflect actual work responsibilities.

## Weekly Cleanup Zones

Nine locations where clutter accumulates:

### 1. Local Downloads
**Path:** `~/Downloads`

Typical contents: PDFs, screenshots, installers, random files from web.

**Action:** File to Work/ or delete.

### 2. iCloud Downloads
**Path:** `~/Library/Mobile Documents/com~apple~CloudDocs/Downloads`

Typical contents: Screenshots (iOS sync), files from iOS apps.

**Action:** File to Work/ or delete. Check for screenshots that should go to projects.

### 3. iA Writer Strays
**Path:** `~/Library/Mobile Documents/27N4MQEA55~pro~writer/Documents`

**Structure:**
```
Documents/
└── Diary/          # Structured diary entries (leave alone)
└── [stray files]   # Waifs outside Diary folder
```

**Action:** Anything outside `Diary/` is a stray. File or delete.

### 4. Desktop (iCloud-synced)
**Path:** `~/Desktop`

iCloud Desktop & Documents sync enabled. Files may be cloud-only (evicted from local storage).

Typical contents: Screenshots, drag-dropped files, temporary staging.

**Action:** File to Work/ or delete. Most Desktop items are transient.

**iCloud voodoo:** Files can appear as stubs (cloud-only, download on access). When bulk processing, may trigger downloads. The `.localized` file is system-managed - leave it.

### 5. Work Folder Root
**Path:** `~/Library/CloudStorage/GoogleDrive-*/My Drive/Work`

Should only contain CLAUDE.md (project instructions) - nothing else belongs at root level.

Typical contents: Stray files that missed proper subfolder, CLAUDE.md.

**Action:** CLAUDE.md stays. Everything else gets filed into Projects/, Areas/, Resources/, or Archive/.

### 6. My Drive Root (Inbox)
**Path:** `~/Library/CloudStorage/GoogleDrive-*/My Drive`

True inbox per PARA. Where `doc.new` creates files, where shared docs land, where quick-captured items accumulate.

Typical contents: New Google Docs, shared files, unsorted captures.

**Action:** Triage into Work/ subfolders or delete. This is the primary inbox to keep clear.

### 7. My Drive Temp (Claude Staging)
**Path:** `~/Library/CloudStorage/GoogleDrive-*/My Drive/Temp`

Nominated temp folder for Claude operations — file uploads, MCP staging, transient outputs.

Typical contents: PDFs uploaded for analysis, intermediate files, staged outputs.

**Action:** Delete everything. If something needs keeping, it should have been filed properly after the session that created it. This folder should be empty between sessions.

### 8. Work Inbox (iCloud)
**Path:** `~/Library/Mobile Documents/com~apple~CloudDocs/Work Inbox/`

Quick capture landing zone from iOS Shortcuts. Subfolders organize by type:

```
Work Inbox/
├── Voice Transcripts/   ← Whisper output from JPR recordings
├── Meeting Notes/       ← iOS Shortcut meeting captures
└── Quick Notes/         ← General quick capture
```

Naming: `YYYY-MM-DD [HH-MM] Title.md`

**Action per subfolder:**
| Subfolder | Triage |
|-----------|--------|
| Voice Transcripts/ | Review, extract actions, file to Meeting Notes/ or delete |
| Meeting Notes/ | Copy to Google Drive Meeting Notes/{year}/, keep original |
| Quick Notes/ | File to PARA, extract to Todoist, or delete |

### 9. JPR Voice Recordings
**Path:** `~/Library/Mobile Documents/iCloud~com~openplanetsoftware~just-press-record/Documents/`

Just Press Record voice memos (m4a/mp3). Organized by date: `YYYY-MM-DD/HH-MM-SS.m4a`

**Action:**
1. Transcribe with `~/.claude/scripts/transcribe-jpr.sh --new`
2. Review transcript in Work Inbox/Voice Transcripts/
3. Audio moves to `Processed/` subfolder automatically

## When Filing Signals Deeper Patterns

File organization chaos often symptoms deeper behavioral patterns. Surface these signals during weekly review:

| Signal | Pattern | Response |
|--------|---------|----------|
| Downloads >100 files OR >2 weeks neglect | **Execution-Without-Reflection** | "This backlog suggests rushing through tasks without processing. Want to check patterns?" |
| Multiple "temporary" folders >1 month old | **Scope Creep** | "These temp folders suggest projects expanding beyond original scope." |

**Coordination pattern:** Filing is Phase 1 of weekly review. After tidying, **todoist-gtd** skill handles pattern reflection (Phase 3).

Don't skip the pattern check just because you tidied the files.

## Processing (Before You File)

Processing means reading and extracting — not just moving.

### Processing Checklist (Meeting Notes)

Before filing ANY meeting note:

- [ ] **Read the content** — Don't just look at the title
- [ ] **Extract actions (mine)** — Concrete next steps I need to take
- [ ] **Extract waiting-fors** — `NAME: What they owe me` format
- [ ] **Note calendar items** — Dates to add to calendar
- [ ] **Quality check: empty?** — Delete (with confirmation)
- [ ] **Quality check: misnamed?** — Rename to match content
- [ ] **THEN move to destination**

**Mandatory extraction:** Cannot file meeting notes until actions are extracted. This is not optional.

### Processing Other File Types

| File Type | Processing Required |
|-----------|---------------------|
| Meeting notes | Full checklist above |
| Quick notes | Check for actions, then file or delete |
| Voice transcripts | Read transcript, extract actions, then file or delete |
| PDFs/docs | Understand content for proper filing location |
| Screenshots | Determine project context, file or delete |

## Action Extraction Workflow (Sublime Loop)

When processing multiple meeting notes with actions:

### 1. Extract to Temp File

Create structured sections:

```markdown
## Waiting For
NAME: What they owe me

## Ping
Quick actions for me to do

## Calendar
Date - Event to add
```

### 2. Open for Review

```bash
open -a "Sublime Text" /tmp/meeting-actions.md
```

### 3. Tell User

"Edit, save, let me know when ready"

User fixes:
- Names (Susie → Susan)
- Stale/irrelevant items (delete)
- Vague actions (clarify)

### 4. After Save

Read file, add to Todoist via todoist-gtd skill.

**Why this works:**
- User sees all actions at once
- Catches errors before they hit Todoist
- Disambiguation happens in familiar editor
- Claude processes clean, user-approved list

## Meeting Source Discovery

At weekly review start, check beyond the 9 zones.

### Standing Docs (Always Check)

**Prompt first:** "Any changes to this standing docs list? (Add/remove as your meetings evolve)"

| Doc | ID |
|-----|-----|
| MIT: Start the Week | `19NqCEPMSOdIdouFqVaHZVA-GyJmkmPjLGyNknTJpU3M` |
| Client Reference | `1owM77DjcTwHjrTfHs9zEfQfNyeFfmtheokBGvkPD8II` |
| Supplier Reference | `1b2vaWpGx-fUj48vuQ0KmIirDu5loodC7F4j3XGV4_DU` |
| Notes – WEEKLY senior team catch up | `1B5X-MrSzldgFElppNlk4s_PDgpIJ0H5Sul74uqQnTs8` |
| TVA/MP/ITV Status | `1tCuph3QWiObYImjb-7nfPPI2Q-D2RjAtou-dceHa45c` |

### Calendar-Linked Docs

**Prompt:** "What recurring meetings had sessions this week? Any have running docs I should check?"

(Future: mise calendar integration to auto-discover — see mise-en-space backlog)

## Filing Decisions

### "Where does this go?"

| If it's... | Put it in... |
|------------|--------------|
| Active project artifact | Projects/{relevant project}/ |
| Meeting notes | Meeting Notes/{year}/ |
| Reference PDF/doc | Resources/Larder/ or relevant Resources subfolder |
| Completed project | Archive/ |
| Ongoing area doc | Areas/{relevant area}/ |
| Screenshot for project | Projects/{project}/ or delete |
| Random download | Delete or file if valuable |

## Personal Documents Zone

**Path:** `~/Documents/Personal/`

Personal documents with PARA structure (no numbering, aligned with Work folder):

```
Personal/
├── Projects/     # Active personal projects
├── Areas/        # Ongoing personal responsibilities (e.g., Ash)
├── Archive/      # Completed/dormant
└── CLAUDE.md     # Project instructions
```

**Note:** No Resources folder - personal "resources" live in Areas as active things.

**Action:** Personal admin docs (scanned letters, etc.) go to appropriate Area.

## Team Shared Drives

**Detection:** `ls "$GDRIVE/Shared drives/"`

Team shared content lives in Shared Drives (not personal My Drive):
```
Team Shared Drive/
├── Team Reference/      # Team-specific reference material
├── Shared Reference/    # Cross-team reference material
└── External Reference/  # External/industry reference
```

**Action:** Team reference materials (shared tools, analytics) go here, not personal Drive.

## Filing Workflow

### Zone Check (Do This First)

At session start, **explicitly check ALL zones** - don't skip any:

```bash
GDRIVE=$(ls -d ~/Library/CloudStorage/GoogleDrive-* 2>/dev/null | head -1)
echo "=== 1. Downloads ===" && ls ~/Downloads | head -5
echo "=== 2. iCloud Downloads ===" && ls ~/Library/Mobile\ Documents/com~apple~CloudDocs/Downloads 2>/dev/null | head -5
echo "=== 3. iA Writer ===" && ls ~/Library/Mobile\ Documents/27N4MQEA55~pro~writer/Documents 2>/dev/null | grep -v Diary | head -5
echo "=== 4. Desktop ===" && ls ~/Desktop 2>/dev/null | head -5
echo "=== 5. Work root ===" && ls "$GDRIVE/My Drive/Work" 2>/dev/null | grep -v "^CLAUDE.md$" | head -5
echo "=== 6. My Drive root ===" && find "$GDRIVE/My Drive" -maxdepth 1 -type f ! -name ".DS_Store" 2>/dev/null | head -5
echo "=== 7. My Drive Temp ===" && ls "$GDRIVE/My Drive/Temp" 2>/dev/null | head -5
echo "=== 8. Work Inbox ===" && find ~/Library/Mobile\ Documents/com~apple~CloudDocs/Work\ Inbox -type f ! -name ".DS_Store" 2>/dev/null | head -5
echo "=== 9. JPR Voice ===" && find ~/Library/Mobile\ Documents/iCloud~com~openplanetsoftware~just-press-record/Documents -name "*.m4a" -o -name "*.mp3" 2>/dev/null | grep -v "/Processed/" | head -5
```

Report which zones have content before starting. Don't skip zones that look empty - verify each one.

### Per-File Process (5 Steps)

When triaging files, follow this 5-step process:

1. **Process** - Read content, extract what matters
   - For meeting notes: extract actions, waiting-fors, calendar items (see Processing section)
   - For other files: understand content for proper filing
   - Use Sublime Loop for batch processing multiple notes
   - For text files: Claude reads directly
   - For binary/GUI files: `open -a "Sublime Text" file.md` or `open file.pdf`
   - **Office files in Google Drive (.docx, .xlsx, .pptx):** These open in native apps (Word, Excel, PowerPoint), not browser. To open in Google Docs instead, fetch the Drive URL via mise and open that:
     ```bash
     # Don't do this (opens in Word):
     open "My Drive/document.docx"

     # Do this instead (opens in browser):
     # 1. Get file ID via mise search
     # 2. open "https://docs.google.com/document/d/FILE_ID"
     ```
     Or flag to user: "These .docx files will open in Word — want Drive URLs instead?"

2. **Rename** - Adjust filename to match actual content
   - Generic names like "Untitled" or "Copy of..." need proper names
   - Use descriptive names: `Simon McCarthy - Pricing Power Proposal`

3. **Split** - If multiple topics in one file, separate them
   - Open in GUI for user review
   - Create new files for distinct topics

4. **Capture** - Outstanding actions → Todoist
   - **Before adding:** Search for existing/completed tasks to avoid duplicates
   - Use todoist-gtd skill (invoke with `Skill(todoist-gtd)`) for Todoist operations
   - Link description to filed document

5. **File** - Move to correct location (or delete)
   - See "Filing Decisions" section for where things go
   - **After filing Meeting Notes:** Extract to memory for searchability
     ```bash
     # Re-scan to pick up the filed document
     cd ~/Repos/claude-mem && uv run mem scan --source local_md

     # Extract entities from the filed document
     SOURCE_ID="local_md:$(basename "$DEST_PATH")"
     cd ~/Repos/claude-mem && uv run mem extract "$SOURCE_ID" 2>/dev/null || true
     ```

## Archive vs Reference

**The test:** "Will I search for this to USE it, or just to REMEMBER it?"

| Answer | Location | Examples |
|--------|----------|----------|
| **USE it** | Resources/ | Methodology docs, templates, how-to guides |
| **REMEMBER it** | Archive/ | Completed project artifacts, old meeting notes |

**Reference** = I'll need this again to do work
- Research PDFs that inform decisions
- Templates I reuse
- Product specs, methodology docs

**Archive** = This was a thing, now it's done
- Completed project artifacts
- Superseded versions
- Historical records

## Common Filing Patterns

### External doc → Google Doc → Project folder

When filing a document that needs to live in a Project:

1. Read content (use docx skill for Word docs)
2. Create Project folder if needed: `{Area} - {Outcome statement}`
3. Convert to Google Doc: `mcp__mise__create`
4. Delete original file
5. Update any Todoist task that references it with new link

### Todoist task linking

When a filed document relates to a Todoist task, use the todoist-gtd skill to search and link. Canonical document location should be in Todoist task description.

## Weekly Review Integration

**Note:** Weekly review is a three-phase workflow orchestrated by **todoist-gtd** skill:
1. **Filing** (this skill) — Clear cleanup zones
2. **Outcomes Review** (todoist-gtd) — Check outcome health
3. **Pattern Reflection** (todoist-gtd) — Freedom score, pattern interrupts

During weekly review, check all 9 cleanup zones:

1. **~/Downloads** - File or delete
2. **iCloud Downloads** - File or delete
3. **iA Writer strays** - File or delete
4. **~/Desktop** - File or delete (mind iCloud stubs)
5. **Work/ root** - Only CLAUDE.md should remain
6. **My Drive root** - Triage into Work/ or delete
7. **My Drive Temp** - Delete everything (Claude staging folder)
8. **Work Inbox** - Triage subfolders (Voice Transcripts, Meeting Notes, Quick Notes)
9. **JPR Voice** - Transcribe new recordings, review transcripts

## Anti-Patterns

| Pattern | Problem | Fix |
|---------|---------|-----|
| File without reading | Actions lost | Mandatory processing first |
| Bulk move meeting notes | Waiting-fors vanish | Use Sublime Loop, extract first |
| Skip running docs | Miss recurring meeting actions | Check standing docs + calendar prompt |
| Skip zones during weekly review | Clutter accumulates | Check ALL 9 zones explicitly |
| Keep "just in case" | Folders bloat | Delete liberally, most things don't need keeping |
| File without renaming | Untitled docs pile up | Rename to content-descriptive name first |
| Skip Todoist search before adding | Duplicate tasks | Always search for existing task first |

## Remember

**File promptly, delete liberally.** Most downloads don't need keeping. Screenshots are usually transient. When in doubt, delete.

**Projects complete, Areas don't.** If you're wondering "is this done?", it's probably a Project. If you're wondering "is this still my responsibility?", it's an Area.

## ~/Repos Naming Convention

Development tools live in `~/Repos/` with prefix conventions:

| Prefix | Contains | Example |
|--------|----------|---------|
| `claude-suite` | Core session skills | session-opening, beads, filing |
| `skill-*` | Specialized skills | `skill-itv-styling` |
| `mcp-*` | MCP servers | `mcp-workspace` |
| `infra-*` | Infrastructure | `infra-openwrt` |
| `claude-*` | Claude experiments | `claude-memory` |

**Skills:** Symlinked to `~/.claude/skills/`. Core skills in `claude-suite/skills/`, specialized skills in individual repos, skills with tooling co-located with their infrastructure.