check-in

Files (1)

• SKILL.md

SKILL.md (first lines)

Process Friday leadership team check-in notes from messy raw notes into polished, consistently formatted documents.

Arguments: $ARGUMENTS

## Base Path

```
$ONEDRIVE_WORK/Meeting Notes/Friday Check-In/
```

Subdirectories:
- `Raw/` — original messy DOCX files (upload new notes here)
- `Cleaned/` — polished DOCX output
- `PDF/` — PDF exports

Create subdirectories if they don't exist.

## Modes

### 1. Single file: `/check-in path/to/file.docx`
Process one file through the full pipeline.

### 2. Batch: `/check-in --all`
Process all unprocessed .docx files in the base directory:
- Enumerate all .docx files in the `Raw/` subdirectory
- **Skip non-check-in files**: Capital Project Meeting.docx, Data Dashboard Preparation.docx, Electric Bus FAQ.docx, Electric Bus Points.docx, Friday Check.docx, $AUTHOR_NAME.docx, Operations Team Meeting Notes Sept 6.docx
- Skip files that already have a corresponding `Cleaned/Friday Check-Ins - YYYY-MM-DD.docx`
- Parse dates from filenames (see Date Parsing below)
- Process in chronological order
- Print progress: "Processing 7/33: Feb 20 2026..."
- Regenerate tracker at the end

#### Agent Teams for Batch Processing

When processing 5+ files, use agent teams for parallel processing:

1. **Create the team**: `TeamCreate` with `team_name: "check-in-batch"`.
2. **Enumerate and filter files first** (skip list, already-processed check), then parse dates.
3. **Create one task per file** via `TaskCreate`. Include the file path, parsed date, and output paths in the task description.
4. **Spawn 3-4 teammates** via `Agent` tool with `team_name: "check-in-batch"`, `subagent_type: "general-purpose"`, named `processor-1` through `processor-4`. Launch all in a single message.
5. Each teammate checks `TaskList` for available tasks, claims one via `TaskUpdate`, processes it through Steps 0-3 (safety check, read, AI cleanup, write cleaned DOCX), marks it completed, then claims the next available task.
6. **Coordinator monitors via `TaskList`**. When all file tasks are complete, collect all extracted action items, regenerate the Action Item Tracker, batch-export PDFs, then shut down teammates and `TeamDelete`.

For fewer than 5 files, process sequentially without teams.

### 3. Tracker only: `/check-in --tracker`
Regenerate Action Item Tracker.xlsx from all existing cleaned files.

## Department Head Mapping

Use this lookup to expand first names found in the notes:

```python
DEPT_HEADS = {
    "debbie": "[Name] - Tech Purchasing & Administration",
    "karen": "[Name] - School Nutrition",
    "julie": "[Name] - Data & EIS",
    "ellen": "[Name] - Finance",
    "melissa": "[Name] - Human Resources",
    "phillip": "[Name] - Maintenance & Facilities",
    "roger": "[Name] - Custodial Services",
    "chuck": "[Name] - Technology",
    "tracie": "[Name] - Coordinated School Health",
    "kristen": "[Name] - Transportation & Operations",
}
```

## Pipeline (per file)

### Step 0: Safety check

Before writing any cleaned file, check if a `Cleaned/Friday Check-Ins - YYYY-MM-DD.docx` already exists AND has been modified more recently than the corresponding raw file. If so, back it up to `~/Documents/Friday Check-In Backups/Friday Check-Ins - YYYY-MM-DD.backup.docx` before overwriting. Create the backup directory if it doesn't exist. This prevents losing manual edits.

### Step 1: Read raw text

Run this Python to extract all paragraph text from the source DOCX:

```python
from docx import Document
import json, sys

doc = Document(sys.argv[1])
paragraphs = []
for p in doc.paragraphs:
    paragraphs.append({
        "text": p.text,
        "style": p.style.name if p.style else "Normal",
        "level": p.paragraph_format.left_indent
    })
print(json.dumps(paragraphs, indent=2, default=str))
```

### Step 2: AI cleanup

Read the raw paragraph JSON and restructure it. Rules:

1. **Parse the date** from the filename or content. Standardize to "Month Day, Year" (e.g., "February 20, 2026").

2. **Identify department sections** by looking for standalone paragraphs that match a first name from DEPT_HEADS (case-insensitive). Also handle variations like "Attendee: Karen, School Nutrition" or "Karen (Nutrition)".

3. **Clean up each person's notes**:
   - Fix obvious typos and spelling errors
   - Convert sentence fragments into clean, complete sentences
   - Organize into a bullet hierarchy: major topics as top-level bullets, supporting details as sub-bullets
   - Maintain third-person perspective ("Karen reported..." not "I reported...")
   - Preserve all factual content, do NOT add information that isn't in the original
   - Keep proper nouns, acronyms, and specific numbers exactly as stated

4. **Extract action items** from each person's section. An action item is any commitment, deadline, follow-up, or decision that requires future action. For each, capture:
   - Person (full name)
   - Department
   - Action item description (clean sentence)
   - Due date (if mentioned, otherwise blank)

5. If a file has duplicate content (same meeting notes appearing twice with different formatting), merge them into one clean version.

### Step 3: Write cleaned DOCX

Run Python to generate the formatted document. Use this structure:

From jasonbhorne/claude-code-skills, MIT.

More from jasonbhorne/claude-code-skills