User Stories

Description

Format specification for writing user stories that document terminal interactions.

Format

Each user story is a single markdown file with the following structure:

Title (H1)

The name of the user story.

Description (H2)

A brief 2-4 sentence overview of what the user story demonstrates.

Transcript (H2)

A text code block showing the terminal output exactly as it appears to the user.

• Use > prefix for user input
• Model responses appear without prefix
• Tool calls and results are shown as the user sees them

> user input here

model response here

Walkthrough (H2)

A numbered list explaining step-by-step what happens during the interaction. Keep descriptions concise. Reference detailed JSON in the Reference section using markdown links:

2. Client builds request (see [Request A](#request-a))

Multiple Transcript Blocks

If the story uses /clear or /rewind to change history, use multiple transcript/walkthrough pairs interleaved:

## Transcript

```text
> first interaction

Walkthrough

1. Step one
2. Step two

Transcript

> /clear
> second interaction after clear

Walkthrough

1. Clear command resets history
2. New interaction begins

### Reference (H2)

Full JSON payloads referenced in the walkthrough. Use H3 headings with anchor-friendly names:

```markdown
## Reference

### Request A

```json
{ ... full JSON ... }

Response A

{ ... full JSON ... }

## Output Location

User stories are stored in `cdd/user-stories/`:

cdd/ ├── README.md # High-level description of release features ├── user-stories/ │ ├── README.md # Index/overview of user stories │ ├── story-1.md # Individual user story (user actions and system responses) │ ├── story-2.md │ └── ... └── ... # Other files/directories (ignored by this skill)

**Note:** The `cdd/` directory may contain other files and directories beyond those listed above. This skill only works with `cdd/user-stories/` and may reference `cdd/README.md`. Other content is permitted and will be ignored.

- `cdd/README.md` - High-level description of release features
- `cdd/user-stories/README.md` - Index/overview with brief descriptions
- Individual story files - User actions and system responses with transcript and walkthrough

Update the README.md as you add user stories to keep the overview current.

## Rules

- One user story per document
- Assume empty message history at start (only system prompt exists)
- Keep model responses short - long responses make discussion harder
- Show exactly what the user sees in transcripts
- Put implementation details (JSON payloads) in Reference section
- Use sequential naming for references: Request A, Response A, Request B, etc.