CO-STARS prompt builder

costars.forge_

Better prompts. Less back-and-forth.

Turn your brief into a structured CO-STARS prompt with costars.forge. Define your goal, audience, constraints, response format and steps up front to help ChatGPT, Claude or Gemini work with fewer missing details. This can reduce clarification turns, saving time and LLM tokens.

Savings and response quality depend on the task and model; results aren't guaranteed. Check the answers you receive.

Start building for free View saved prompts

prompt.costars

# Context
I'm planning a small surprise birthday dinner for my sister. She's gluten-free, dislikes noisy restaurants, and her birthday is next Saturday.

# Objective
Suggest three suitable restaurants near her apartment, with one reason each fits and one possible downside.

# Style
Plain-language bullet points, no marketing fluff.

# Tone
Helpful and practical, like a well-organized friend.

# Audience
Me — a busy sibling who doesn't know the neighborhood well.

# Response format
Markdown list: name, address, price range, why it works, possible downside.

# Steps
1. Check every option against the dietary, noise, date, location, and budget constraints.
2. Rank the three strongest matches and explain each trade-off.
3. Verify the final list includes every required field before answering.

More clarity up front. Fewer corrections later.

Missing details often mean another turn explaining what you meant. CO-STARS organizes Context, Objective, Style, Tone, Audience, Response format and Steps before you send your prompt. That structure can reduce repeated explanations and corrections — helping you spend less time and fewer LLM tokens on back-and-forth, without guaranteeing an accurate answer.

Try the builder

One-liner prompt

Vague one-liner the model has to interpret
Missing constraints, audience, or output shape
Extra time and tokens spent clarifying missing details
More room for the model to make assumptions

CO-STARS prompt

Seven structured sections with requirements up front
Context, Objective, Style, Tone, Audience, Response format, Steps
Can reduce clarification turns and save time and LLM tokens
Reusable template you can refine and store

Built for prompt work, not for demos

Guided, interactive stages

Give the raw problem context. The builder walks you through Context → Objective → Style → Tone → Audience → Response format → Steps, one question at a time.

Suggestions, clearly labelled

Every AI-generated idea appears in a dashed suggestion panel marked as the app's suggestion. Use it, append it, or ignore it — your words stay yours.

Saved to your library

Every prompt is stored to your account, searchable and re-editable, so you can iterate on it later instead of starting over.

Dark and light, done properly

A terminal-flavoured theme built on semantic tokens, so every colour — text included — flips correctly between modes.

The seven stages

CO-STARS keeps a prompt complete without making it bloated. Each letter is one focused question.

C
Context
Background the model needs before it can be useful.
O
Objective
The single concrete task the model must perform.
S
Style
Writing craft: structure, vocabulary, level of detail.
T
Tone
Emotional attitude of the response.
A
Audience
Who reads the output and what they already know.
R
Response
Exact shape of the output the model must return.
S
Steps
The sequence of steps the model should follow to get there.

Open the builder