How should an AI SRE end its run when failed deploys turns out to be nothing?

Tares lets an AI SRE end that run with the conclude tool set to no_op, which records no finding. When failed deploys is a real problem, the agent concludes with a finding instead: a verdict about a named entity, a headline and a next step. The run then ends on purpose either way, and nothing noisy lands in your project.

How it works

An agent that ends with a conclusion gets the conclude tool. It ends a run in one of two ways, described in How a run ends:

• no_op: there is nothing to report. No finding is recorded. Use this when the check on failed deploys finds that nothing is wrong.
• A finding with a verdict about a named entity, plus two more 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 stays empty when nothing needs doing.

The headline and next step are what the project’s Overview shows on each card, and agents are told to give both. If a run ends without them, Tares takes the headline from the note’s heading, bold lead or first sentence. It takes the next step from the text after “Next step:”, “Next:” or “Recommendation:”.

When you turn on Ends every run with a conclusion (concludes: true), the agent always gets the conclude tool, whatever its prompt says. Tares adds the instructions for it, so your prompt only describes the job. If a run stops without concluding, Tares asks once more with conclude as the only choice. With the setting off, the agent’s last message is its finding, with no verdict.

You can also define Verdicts it can give (verdicts). Each is your own word with a note on when to give it. When you set a list:

• The tool accepts only those words and tells the model what each one means.
• A finding must carry one of them.
• 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. The word no_op is always available and never goes on your list.

Set it up

1. Open the agent form and go to the How it ends section.
2. Tick Ends every run with a conclusion. In YAML this is concludes: true.
3. Under Verdicts it can give, add the words you want for real findings. Give each one a short note on when to use it. You can add up to 10, each one word made of letters, digits or dashes, and each listed once. Skip this step if any verdict, or none, is fine.
4. Write a prompt that only describes the job, for example checking failed deploys and judging whether they point to a problem. Tares adds the conclude instructions for you.
5. Check the agent’s summary, which shows how it ends.

The same setup in YAML:

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

With this agent, a run that finds failed deploys harmless ends with no_op. A run that finds one entity that looks like a problem ends with a finding. That finding carries the verdict investigate, a headline of up to 100 characters and a next step of up to 300 characters.

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. Handoffs key off these words, so choose them with that in mind.

What you get

• A clear end to every run. The agent concludes on purpose instead of trailing off with a loose last message.
• No finding for a quiet result. When failed deploys turns out to be nothing, no_op leaves your project free of empty or misleading entries.
• Cards you can scan. A real finding shows a headline in the reader’s terms and a next step a person can act on, or an empty next step when nothing needs doing.
• Consistent verdicts. With a verdict list, findings use your own words, and a wrong or missing word is corrected inside the same run.
• A second chance. If the agent stops without concluding, Tares asks once more with conclude as the only choice.

For the wider picture of running an agent as an SRE, see AI SRE.

When something else fits

• If you want the agent’s last message to be the finding, leave the conclusion setting off. Findings then carry no verdict.
• If you do not need fixed verdict words, skip the list. A finding can then carry any verdict, or none.

Related