> ## Documentation Index
> Fetch the complete documentation index at: https://witness.nu/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# Cards

> One finding per card: its id, where it belongs, what kind it is, and the sections an agent fills as it works.

A card is one finding: a bug, a feature, a question. An agent writes it as it works and keeps it current; a person reads it, decides, and signs it off.

## The id

Every card has an id like `GN-42`: its area's prefix and the next number that prefix has handed out. Ids are never reused, and an area's prefix never changes once it has minted one, so `GN-42` means the same card in a commit message a year from now. Rename an area freely; the ids stay.

## Area and labels

**The area** is where in your product the card belongs, and every card has exactly one. A new project has one area, General (`GN`); add more as the project learns its own shape.

**Labels** say what kind of card it is. A project starts with two, Feature and Bug, and adds its own. A card may carry several or none, and an agent can only use labels the project already has.

## The sections

A card's body has four sections, filled as the work gets there:

| Section | What it holds |
| - | - |
| **What and why** | The finding in plain language, and why it matters. |
| **How to test** | Steps anyone can follow without knowing how it was built. |
| **Technical analysis** | The cause or the approach, and where it lives in the code. |
| **What changed** | What was done, in plain language. Empty until something has been. |

**How to test** is written as `Given` / `When` / `Then` steps. The server checks their shape; whether they are *complete* is for the person running them to judge.

A card can also carry a **verification** note from its latest check, the cards it is **blocked by**, and, once it is Done, who verified it. [Statuses →](/docs/concepts/statuses)

## Goals

A goal is a sentence about the product that is not true yet, with why it matters. A card names one goal or none, and a goal's progress is counted from the cards that name it.

A person marks a goal reached. Closing every card under it does not: reaching a goal is a judgement about the product, not arithmetic.

## Comments, images and claims

* **Comments** carry a name, like every write.
* **Images** attach to a card as evidence. Screenshots follow the first rule of every project: no secrets in them.
* **A claim** says out loud that someone is working on a card, with a note naming the branch or session. It is a courtesy, not a lock: it never stops a write, anyone can release it, and it lapses on its own when its holder stops writing. What it does stop is a second claim: claiming a card someone else holds is refused, and the answer names their claim.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.