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

# Writing effective prompts

> Get better answers from Mesa by writing specific, well-grounded prompts.

Mesa performs best when it knows exactly what you want, what materials to rely on, and what the output should look like. A few habits make the difference between a generic answer and one you can use directly in your work.

## Be specific about what you want

Vague questions get vague answers. Name the jurisdiction, document type, parties, and time period whenever they matter.

| Instead of                               | Try                                                                                                                                     |
| ---------------------------------------- | --------------------------------------------------------------------------------------------------------------------------------------- |
| "What's the law on financial reporting?" | "Summarize Delaware's financial reporting obligations for public companies."                                                            |
| "Review this contract."                  | "Identify every indemnification and limitation-of-liability clause in this agreement and flag terms that favor the counterparty."       |
| "Help me with this memo."                | "Draft a two-page memo for a senior partner analyzing whether these facts support a breach of fiduciary duty claim under New York law." |

## Give Mesa context

Tell Mesa who the output is for and what role it should take. An answer written for a client reads very differently from one written for internal counsel.

> Write this as if you are a corporate associate briefing a client with no legal background. Keep it under one page and avoid jargon.

## Describe the output you need

If you want a memo, a table, a redline summary, or a bullet-point checklist, say so up front. Mesa will match the structure you describe.

> Return the results as a table with columns for section number, clause type, and risk level.

## Ground Mesa in your documents

Mesa gives more reliable answers when it works from your actual materials instead of general knowledge:

* Type `@` in the chat box to tag files you've saved to Mesa. See [Tagging files](/tagging-files).
* Click the paperclip icon or drag and drop to attach one-off documents. See [Attachments](/attachments).

## Ask one primary question per message

When a message bundles several unrelated questions, the answer to each gets shallower. For focused, thorough answers, send one core question at a time and follow up in the same conversation.

## Ask for completeness when you need it

A normal request finds the best answer. When you need *every* matching item (for a dataset, compliance checklist, or issue spotter), explicitly request an exhaustive search:

> Run an exhaustive search for all indemnification provisions in these contracts.

<Tip>
  Use the exact words "exhaustive search" so the agent knows to catalogue every match rather than summarize the best ones. See [Exhaustive search](/exhaustive-search).
</Tip>

## Save what works

Once you find phrasing and formats that work for you, you don't need to retype them:

* **[Skills](/skills)** save task-specific instructions you reuse for recurring work, like a contract review checklist or a memo template.
* **[Preferences](/preferences)** apply standing instructions to every conversation, like your practice area, default jurisdiction, and preferred format.

## Related articles

<Card title="Text editor" href="/editor" horizontal>
  Learn the chat box's shortcuts and file-referencing features.
</Card>

<Card title="Models" href="/models" horizontal>
  Pair a well-written prompt with the right model and effort level.
</Card>

<Card title="Skills" href="/skills" horizontal>
  Turn your best prompts into reusable instructions.
</Card>
