# SKILL: &lt;short, verb-first name&gt;

> **What this is:** a one-page, reusable "recipe" you paste into any chatbot (ChatGPT, Gemini, Claude - whichever you have) to reliably repeat a real task. Fill the seven sections top to bottom. The order is deliberate: you describe the result **before** you write the prompt. That's [Way #1: Describe before you generate](../ways-of-working.md).
>
> Keep it to one page. Delete these quote-blocks before you print.

---

## 1. Goal - what this Skill does, and when to use it
*One or two sentences. What real task does this replace? When would you reach for it?*

> e.g. "Turn my messy Monday call notes into a consistent status update. Use every week before the programme report."

## 2. Inputs - what you paste in for it to work
*List each thing the Skill needs.*

- Input 1:
- Input 2:

> **PRIVACY CHECK** ([Way #3](../ways-of-working.md)): before you paste anything, list what to remove or disguise.
> e.g. *Replace real names with Person A/B. Replace client/vendor names with Client 1. Remove any figure marked confidential. No ID numbers.*

## 3. Output - what "good" looks like *(write this BEFORE the prompt)*
*The exact shape of the result: sections, columns, length, tone. This is the heart of spec-first - if you can describe it here, the AI can produce it.*

> e.g. "A table with columns: Summary | Key decisions | Risks (rated Red/Amber/Green) | Actions (Owner, Due). Max 8 rows. Neutral, factual tone."

## 4. Instructions / Prompt - the exact text to paste
*Write it so anyone can copy it word-for-word. Put the bits you swap each time in [SQUARE BRACKETS]. Include the output shape from section 3, and an anti-guessing instruction.*

```
You are helping me [do the task].
Here is [the input] for [PROJECT / SUBJECT].
Produce [the exact output from section 3].
If something is missing, write "TBC" - do not guess or invent.
Keep it [tone].
Input: [PASTE ANONYMISED INPUT]
```

## 5. Worked Example - one real run, anonymised
*Proof it works, and a yardstick to compare against next time.*

> **Input:** *(a short, anonymised sample)*
>
> **Output it produced:** *(the actual good result)*

## 6. Guardrails - how to stay safe and sane
- **Privacy** ([Way #3](../ways-of-working.md)): never paste &lt;X&gt;. Always anonymise &lt;Y&gt;.
- **Human check** ([Way #2](../ways-of-working.md)): before you trust it, verify &lt;these 1–2 things&gt; - AI may invent owners, dates, or figures.
- **Limits:** don't use this for &lt;out-of-scope cases&gt;.

## 7. Reuse notes - so it isn't lost
- **Owner:**
- **Last updated:**
- **Lives in:** *(where your team keeps it - a shared drive / wiki page, so others can reuse it - [Way #5](../ways-of-working.md))*

---

*Built at Innovation Day. The five ways of working: [ways-of-working.md](../ways-of-working.md).*
