What does the conclude tool let an agent do, and what happens if a run stops without calling it?

Tares gives an agent the conclude tool to end a run on purpose, either with no_op when there is nothing to report or with a finding that carries a verdict, a headline and a next step. With Ends every run with a conclusion on, a run that stops without concluding gets one more prompt where conclude is the only choice. With it off, the agent’s last message is the finding, with no verdict.

How it works

The conclude tool gives a run two ways to end, both on purpose:

• no_op: there is nothing to report. No finding is recorded.
• A finding with a verdict: the verdict is about a named entity. The finding also carries two fields.
  • headline: one line of up to 100 characters. It says what was found in the reader’s terms.
  • next_step: up to 300 characters. It says what a person should do, and it is empty when nothing needs doing.

The headline and the next step are what the project’s Overview shows on each card, and agents are told to give both.

The Ends every run with a conclusion setting (concludes: true) controls what happens around the tool:

• On: the agent always gets the conclude tool, whatever its prompt says. Tares adds the instructions for using it, so your prompt only describes the job. If a run stops without concluding, Tares asks once more with conclude as the only choice.
• Off: the agent’s last message is its finding, with no verdict. An agent whose prompt names conclude still gets the tool. Tick the box to make concluding required.

When a run ends without a headline or next step, Tares fills them in. The headline comes from the note’s heading, bold lead or first sentence. The next step comes from the text after “Next step:”, “Next:” or “Recommendation:”.

You can also restrict the verdicts an agent may give. Each verdict is your own word, with a note on when to give it:

• You can define up to 10.
• Each is one word made of letters, digits and dashes.
• Each is listed once.
• None is no_op, because nothing to report is always possible.

With a list, the tool accepts only those words and tells the model what each one means. A finding must carry one. An off-list word or a missing verdict goes back to the agent to fix within the same run. Without a list, a finding can carry any verdict, or none.

This fits into the wider flow described in AI SRE, where agents watch your systems and report what they find.

Set it up

1. Open the agent form and find the How it ends section.
2. Tick Ends every run with a conclusion. This sets concludes: true and makes the conclude tool always available to the agent.
3. Under Verdicts it can give, add your own words if you want a fixed set. Give each one a short note on when to use it.
4. Keep the prompt about the job only. Tares adds the conclude instructions for you.
5. Save the agent. Its summary shows how it ends.

The same setup in YAML looks like this:

agents:
  - name: triage
    trigger: watch-web
    concludes: true
    verdicts:
      - { verdict: investigate, when: one entity looks like a problem and needs a closer look }
    prompt: |
      ...

The Triage (on a schedule) preset comes with the investigate verdict. The Root cause after a handoff preset comes with rca, for when you found what is failing and why, and resolved, for when it was noise or is already over. An agent’s handoffs key off these verdict words, so choose words that match how you want work to move on. The full behavior is on How a run ends.

What you get

• Clean quiet runs. An agent that finds nothing ends with no_op, and no finding is recorded, so your findings list holds only things worth reading.
• Findings you can scan. Each finding has a verdict, a headline of up to 100 characters and a next step of up to 300, so each Overview card says what happened and what to do.
• A second chance to conclude. With the setting on, a run that stops early is asked once more with conclude as the only choice, so you do not end up with a run that just trails off.
• Controlled vocabulary. With a verdict list, off-list words and missing verdicts go back to the agent to fix in the same run, so the words your handoffs depend on stay consistent.
• Shorter prompts. Tares adds the conclude instructions, so the prompt only describes the job.

When something else fits

• Leave Ends every run with a conclusion off when you want the agent’s last message to be the finding as plain text and you do not need a verdict.
• Skip the verdict list when you want a finding to carry any verdict, or none. Use a list when handoffs or readers depend on a fixed set of words.

Related