A prompt is the instruction or question a language model receives. The wording shapes the result: the more specific the task and context, the more accurate the answer.
A good prompt usually covers:
- The task: what needs to get done
- The context: which data to rely on
- The format: how to present the result
- The constraints: what to avoid
In Pachca, you don't need a separate prompt window: @mention the agent in a thread and describe the task in a regular message, like “create a task from this discussion.” The agent picks up the context from the thread itself. To help the agent figure out the Pachca API, add the link dev.pachca.com/llms.txt to your prompt.
For developers
Docs on dev.pachca.com →