# The one-page spec for an agent build (free sample)

> Four questions to answer in ten minutes before a coding agent writes a line, as a SPEC.md template with four rules and a check prompt.

- URL: https://paperchasewebb.ai/bots/marketplace/items/one-page-spec-for-an-agent-build-free-sample/
- Type: Prompt pack
- Category: Vibe coding
- By: PaperChaseWebb
- Works with: Claude Code, Codex, Cursor (named in the listing); Plain .md
- Price: Free USD
- Difficulty: Beginner
- Setup time: 10 min
- Time to result: Ten minutes
- Format: Markdown (.md) file, one-page-spec-for-an-agent-build-free-sample.md
- Version: 1.0, updated 2026-10-26
- License: Personal and commercial use.

## What you end up with

Your agent builds from your decisions, not its guesses, even on a weekend project.

## What you get

- The SPEC.md template
- Four rules
- A check prompt

## Who it is for

Anyone who starts an agent build from a single prompt.

## Tools needed

- Claude Code: reads SPEC.md first
- Codex: reads SPEC.md first
- Cursor: reads SPEC.md first
- You need: Any coding agent

Cost to run: Free.

## Sample prompt: Check the spec (Claude Code, Codex, Cursor)

```
Read SPEC.md as the agent that will build this. List every place you would have to guess, and write the question I need to answer for each. Don't start building.
```

Answer the questions, then start.

## Free preview of one-page-spec-for-an-agent-build-free-sample.md

````markdown
# Free sample: the one-page spec for an agent build

From Spec-First Build Docs by PaperChaseWebb. Before a coding agent writes a line, answer four questions. It takes ten minutes, and it is the difference between an agent carrying out your decisions and an agent making them up.

Save this as `SPEC.md` in your project and tell the agent to read it first.

## Template

```markdown
# SPEC: [project]  (updated [date])

## 1. What we are building, and what we are not
In this version:
- 
Not in this version (make this list longer):
- 

## 2. What it should look and sound like
Rules:
- 
Never:
- 

## 3. What it is made of, and what it must never touch
Stack:
Data lives in:
Never touch:

## 4. What the agent may do alone
Alone:
Stop and ask before:
Never:
```

## Four rules

1. If you can't answer a question in a sentence, figure that out before the agent starts.
2. Anything undecided goes on a list with your name next to it. The agent doesn't decide it.
3. If two readings of the spec are possible, the agent stops and asks.
4. When the project gets a second user, a second builder or a second week, turn each answer into its own document.

## Check prompt

```text
Read SPEC.md as the agent that will build this. List every place you would have to guess, and write the question I need to answer for each. Don't start building.
```

The full pack has the PRD, design system, architecture note and agents guide templates, a project tree for Claude Code, Codex and Cursor, a weekly staleness pa
````

## Worth knowing

- A one-page spec is for experiments. Client and public builds need the full set.

## Questions

**What is in the full pack?** PRD, design system, architecture note and agents guide templates, a project tree for Claude Code, Codex and Cursor, a weekly staleness pass, a nine-area readiness check and 12 prompts.

## How to get it

Free sample. Download it directly at https://paperchasewebb.ai/bots/marketplace/items/one-page-spec-for-an-agent-build-free-sample/, no signup.

Seller: PaperChaseWebb Inc., mgmt@paperchasewebb.com