A prompt is how you tell Zo what to do. The clearer your prompt, the better the result. Zo can read your files, browse the web, run code, and call hundreds of tools, so most of the quality of an answer comes down to how well you describe what you want.Documentation Index
Fetch the complete documentation index at: https://docs.zocomputer.com/llms.txt
Use this file to discover all available pages before exploring further.
What good prompts look like
The three habits below cover most of the gap between a so-so result and a great one.Be specific
Vague prompts get vague answers. Define what “done” looks like, give an example or two, and call out anything Zo should avoid.- Vague: “Summarize this PDF.”
- Specific: “Summarize this PDF in five bullets, each under twenty words. Skip the methodology section. End with one open question I should follow up on.”
Provide context
Type@ to mention specific files, folders, or chats in your workspace. You can also point at tools you want Zo to use, send URLs you’d like it to read, or attach images and screenshots. The more grounded the prompt, the less Zo has to guess.
- “Use @notes/q3-planning.md and @reports/sales to draft a Q3 review email.”
- “Read https://example.com/changelog and tell me what changed since last Tuesday.”
Break down complex tasks
For anything multi-step, ask Zo to plan first, then execute. You can have it write the plan into the chat or save it to a note you both work from.- “Plan a migration from Postgres 14 to 16 for the
services/apifolder. List the steps in order, then stop and wait for me before doing anything.”
Use the Prompt Improver skill
The official Prompt Improver skill rewrites your prompt before Zo runs it, applying the practices above automatically. Click the link to install it on your Zo, then turn it on whenever you want a sharper version of what you typed.Where prompting shows up elsewhere
The same practices apply when you’re writing:- A bio, so Zo has lasting context about you.
- A persona prompt, so a character has a consistent voice.
- A rule, so Zo behaves a certain way in specific situations.
- An automation, so a scheduled task knows exactly what to do without you in the loop.