A simple pattern for clearer prompts

Good prompts tell Babbily three things: what you’re trying to accomplish, what context matters and how you want the answer shaped. You don’t need special phrasing. You need to be specific.

For work you repeat, reuse a consistent prompt structure and save stable context, such as your role, audience, tone and formatting preferences, to memory. Keep task-specific instructions in the prompt itself so you can adjust them each time.

The simple pattern

When you want a reliable answer, write five short lines:

• Goal: what you want.

• Context: what Babbily should know.

• Audience: who it’s for.

• Constraints: tone, length, sources, tools, files or style.

• Format: bullets, a table, a memo, a plan, a checklist, a draft or a chart.

Here’s the pattern filled in for a customer update:

“Goal: Draft a customer update. Context: We fixed the issue and monitoring is stable. Audience: Non-technical customer admins. Constraints: Calm, concise, no blame, include next steps. Format: Email draft with subject line.”

Ask for the shape of the answer

If you don’t specify a format, Babbily chooses one. When you know what you need, say so:

• “Return this as a table.”

• “Give me a 5-step checklist.”

• “Write this as a customer email.”

• “Summarize this as risks, decisions, and next steps.”

• “Show this as a bar chart if the data is clear enough.”

Tell Babbily what to use

Name the source of context when it matters. “Use the attached file.” “Use web search and cite sources.” “Show me photos of the venue.” “Use Gmail to find the latest customer thread.” “Use my Memory Profile only if it helps.” Or the opposite: “Do not use tools for this reply.”

Improve an answer with follow-ups

The first reply is a draft you can shape. Useful follow-ups include:

• “Make it shorter.”

• “Add examples.”

• “Explain the trade-offs.”

• “Turn this into a checklist.”

• “Make it more customer-friendly.”

• “What assumptions did you make?”

• “What should I verify before using this?”

Recipes to start from

Drafting. “Draft a [format] for [audience]. The goal is [goal]. Use a [tone] tone. Include [required details]. Avoid [things to avoid].”

Research. “Research [topic]. Use current public sources, cite them, and return a table with findings, source links, confidence, and open questions.”

File analysis. “Analyze the attached file for [goal]. Return a summary, risks, important details, and recommended next actions.”

Plugin work. “Use [plugin] to find [context]. Summarize [desired result]. Ask before sending, updating, or changing anything.” Babbily asks with an approval card, and only choosing Allow on the card runs the action. A typed reply doesn’t approve it.

Creative work. “Create [image or video] for [audience/use]. Style: [style]. Include [elements]. Avoid [constraints]. Shape: [square, portrait, or landscape].” Babbily uses the Image generation or Video generation tool for this. To edit an existing image or video, attach it and describe the change.

When answers miss the mark

• Restate the goal.

• Add the audience.

• Ask for a different format.

• Give an example of the style you want.

• Say what to ignore.

• Start a new thread if old context is steering the answer.

• Switch models when the task needs a different capability, or choose Auto to let Babbily pick a model for each message.

Where to go next

The task recipes guide builds on this pattern with complete workflows for files, research, plugins, media and translation. When a task needs a particular strength, such as speed, deeper reasoning, files, tools, media or a longer context, the model capabilities page in the docs compares the named models. Or leave the choice to Auto.

For repeat work, save the stable parts of the pattern, such as your role, audience, tone and formatting preferences, to memory. Keep the goal, constraints and format in each prompt so you can change them as the task changes.

Read the full docs ↗