# The Astra-nomical Prompt

A starting brief for **GPT-6 Astra**.

1. Open [The Astra-nomical Prompt](prompts/02-take-this-job.txt) and fill in the brackets with a real job.
2. Supply the files, notes or references it needs.
3. Let it work. Give it useful feedback when you see the result.

The prompt gives Astra an outcome and a finish line. It also tells Astra to make routine decisions and continue independent work while asking a meaningful question. Keep any job-specific limits in the brief.

## Try it

[Weekly client update](examples/weekly-client-status.md) is a fictional exercise with all the source files and an expected draft. [The paper intro](examples/intro-worked-example.md) shows actual creative directions from this lesson, plus a reconstructed brief.

## Optional follow-ups

Use these when you need them:

- [Explain how I work](prompts/01-explain-how-i-work.txt)
- [Change direction](prompts/03-change-direction.txt)
- [Show the work and proof](prompts/04-show-work-and-proof.txt)
- [Improve the next run](prompts/05-improve-next-run.txt)

Unzip the folder and open the files in any text editor. The pack is readable offline. Running the prompt requires access to GPT-6 Astra and the material or tools your job needs.

## Why this wording

The brief follows OpenAI's current Astra guidance on initiative, context and follow-through. Firsthand reports from Matt Shumer, Kilo, Chris Best and Peter Yang support keeping prompts focused while showing the model what matters. This is a starting prompt, not a measured optimum.

- https://developers.openai.com/api/docs/guides/latest-model
- https://somethingbig.ai/astra-review
- https://blog.kilo.ai/p/gpt-6-astra-what-we-learned-previewing
- https://cb.substack.com/p/early-reflections-on-gpt-6-astra
- https://creatoreconomy.so/p/gpt-6-astra-is-both-incredible-and