Back to home
Tasks & conversation

Create a task

Create a task

Aureink’s core strength is understanding natural language. You don’t need to learn complicated steps — describe the outcome you want in a single sentence, and it plans and executes the work on its own, like a colleague, handing back results you can sign off on directly. This page covers how to assign a task and how to add information so the results land closer to what you expect.

1 · Describe your task

In the input box on the Chat home, just type the work you want done. We recommend describing “the result you want,” not “how to do it.”

  • Example: “Turn the customer feedback in this folder into a weekly report, distilling the key issues and follow-up actions.”
  • Example: “Analyze this sales data, find the categories that declined month over month, and generate a chart.”

If you’re not sure where to start, click one of the work mode shortcuts above the input box: Write code, Data analysis, Research, Write docs, Spreadsheet, Automate task, Build website, or Chat mode. Once selected, Aureink organizes its execution to fit that mode.

〔Figure: Chat home with work-mode chips, marking the input box and mode entries〕

2 · Add task context (the + menu)

Click the + on the left of the input box to attach the materials and scope this task needs:

  1. Attach file — reference a local file or image as the object to work on; you can drag it onto the input box or paste a file directly.
  2. Select Skill — assign a Skill to the current task, or type / in the input box to invoke it directly. We recommend enabling only the Skills you need this time, to reduce accidental calls.
  3. Select Agent — bind an Agent to this conversation. This can’t be changed after the first message is sent, so choose before you start.
  4. Workspace — specify the folders Aureink can read and write. It works only within the directories you authorize.
  5. Model — view or switch the model used this time.

〔Figure: The + menu expanded, each item labeled〕

3 · Set the permission mode

Near the input box you can switch this task’s permission mode, which decides whether Aureink needs your confirmation before it performs sensitive operations:

ModeWhen to use
Ask when neededDefault. Asks for your approval before actions such as writing files, running commands, or going online.
Auto within sandboxAutomatically allows actions within what the sandbox permits, reducing interruptions — good for batch tasks you already trust.

If your organization has enabled Enterprise knowledge, you can turn on the switch here to let this task draw on knowledge within your organization’s authorized scope.

4 · Information worth adding up front

To get results right the first time, it helps to spell out the following when you assign the task:

  • Goal — what you ultimately want (a report? a chart? a usable file?).
  • Scope — which files, which time range, and for whom.
  • Format & tone — the structure, style, form of address, or template you expect.
  • Constraints — which files not to touch and what must be kept.

Incomplete information is fine: Aureink will ask you back when needed (structured clarification), and continue once you fill in the gaps.

5 · Task created

Once you send it, the task starts running. You’ll see its thinking trace, the tools it’s calling, and step-by-step progress; complex tasks may split into sub-tasks that run in parallel. At this point you can:

  • Keep asking follow-ups or adding notes in the conversation;
  • View interim outputs in the result panel on the right;
  • Stop generating at any time, or restore file changes from a given step.

〔Figure: Thinking trace and tool calls during task execution〕

Next step: To learn how to collaborate with Aureink mid-execution, see Conversation →

Notes & key reminders

  • File access scope: Aureink only reads and writes the folders you authorize in a workspace; it won’t access other directories on its own.
  • Sensitive operations: Actions such as writing/deleting files, running commands, or going online request confirmation according to the permission mode you chose; high-risk operations have a second confirmation and a countdown lock.
  • Quota usage: Running tasks consumes usage quota; complex tasks and stronger models consume more.
  • Result review: Generated results are for reference — have a human review key content before delivery.

Last updated · July 2026 | Next: Conversation →

July 2026