# Skill Brief — Onboarding

## Your Mission

Build a custom skill that automates the task you did manually in Workshop 1 or 3. Study the example skill in `example-skill/SKILL.md` first, then create your own.

## What to Build

**Skill name:** `stuck-finder`
**Trigger words:** "stuck applications", "onboarding bottleneck", "stuck finder"

**Input:** CSV with Application_ID, Merchant, CDD_Status, Ops_Status, Days_Pending
**Output:** .docx report with bottleneck analysis and follow-up actions

## Steps to Create

1. Run `/skill-creator` in Cowork
2. When asked for a name, enter: `stuck-finder`
3. When asked what it does, describe: "Read application-pile data and produce a  report with priority categorisation"
4. When asked for constraints, include:
   - "Never fabricate data"
   - "Flag items with missing fields as NEEDS MANUAL REVIEW"
   - "Include a disclaimer that output requires human review"
5. Test against `test-input.xlsx` in this folder

## Helper Script

A helper script (`helper-script.sh`) is included. Your skill can call it for calculations:
- **What it does:** Takes days_pending + sla_target → outputs SLA status and urgency
- **How to use:** `bash helper-script.sh <arg1> <arg2>`

## Success Criteria

Your skill passes if:
1. It processes `test-input.xlsx` without errors
2. Output contains a summary table with correct counts
3. Priority categorisation matches the thresholds in the example skill
4. It handles the edge cases in `edge-cases.csv` gracefully (no crashes, flags issues)