What verdict and next step should an AI SRE record for a disk filling up?

Tares lets an AI SRE end a run on a filling disk with the conclude tool, recording a finding with a verdict, a headline and a next step. Use a verdict such as investigate while the disk is filling and needs a closer look, rca once you know what is filling it, and resolved when it is noise or already over. Give the next step as the action a person takes, such as freeing space or growing the volume. If nothing needs reporting, end with no_op and Tares records no finding.

How it works

An agent that ends with a conclusion gets the conclude tool, which ends a run on purpose. It has two outcomes:

• no_op: there is nothing to report, and no finding is recorded.
• A finding: a verdict about a named entity, here the host, volume or mount that is filling up. The finding also carries a headline and a next step.

The two text fields have fixed limits:

• 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 project’s Overview shows both 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:”.

Verdicts are your own words. You can define up to 10. Each is one word made of letters, digits and dashes, listed once, and none can be no_op. 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, and an off-list word or a missing verdict goes back to the agent to fix within the same run. Read the full rules on How a run ends.

Set it up

1. Open the agent form and find the How it ends section.
2. Tick Ends every run with a conclusion (concludes: true). The agent then always gets the conclude tool, and Tares adds the instructions for it, so your prompt only describes the job, such as checking a disk that is filling up.
3. Add the verdicts the agent can give, each with when to give it. For a disk, a small set covers the cases:
   • investigate: one entity looks like a problem and needs a closer look. This is the word the Triage (on a schedule) preset comes with, and it fits a disk whose usage keeps climbing.
   • rca: you found what is failing and why. This is a word the Root cause after a handoff preset comes with. Use it when the agent has traced the growth to a cause.
   • resolved: it was noise, or it is already over. This is the other word from that preset. Use it when usage dropped back after a cleanup or rotation.
4. Write the headline and next step in the prompt’s terms, within the limits above. For example:
   • Headline: Disk on db-1 is filling up
   • Next step: Free space on the data volume or grow it, then check what keeps writing to it
5. Save the agent. The agent’s summary shows how it ends, and its handoffs key off these verdict words.

The agent definition 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: |
      ...

Add more entries under verdicts for rca and resolved in the same shape, each with its own when. If the agent stops without concluding, Tares asks once more with conclude as the only choice. For the wider picture of agents that watch and escalate, start from AI SRE.

What you get

• A finding on the project’s Overview with a verdict, a one-line headline and a next step a person can act on.
• Verdicts you control. With a list, the agent can only pick from your words, so a disk finding never carries a stray label.
• Quiet runs stay quiet. When the disk is fine, no_op records no finding, so the Overview shows only what needs attention.
• Verdicts that other agents can use. Handoffs key off these words, so an investigate finding can pass to an agent that ends with rca or resolved.
• A fallback for incomplete runs. A run that ends without a headline or next step still gets both from its note, and a run that stops without concluding is asked once more.

When something else fits

• If you want the agent’s last message to be its finding with no verdict, leave Ends every run with a conclusion off. The finding then carries no verdict.
• If you do not want to fix a vocabulary, leave the verdict list empty. A finding can then carry any verdict, or none, but nothing checks the word.
• If the disk is fine and there is nothing to say, end with no_op rather than recording a finding.

Related