What verdict and next step should an AI SRE record for a cron job that stopped?

Tares has an AI SRE end its run with the conclude tool, recording a finding for the stopped cron job with a verdict, a one-line headline and a next step. The verdict is one of the words you define on the agent, such as investigate when the job looks like a problem and needs a closer look. If there is nothing to report, the agent calls conclude with no_op and Tares records no finding. See How a run ends for the full behavior.

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: nothing to report, so no finding is recorded. Use it when the cron job ran as expected.
• A finding with a verdict about a named entity, here the cron job that stopped.

A finding also carries two more fields:

• headline: one line of up to 100 characters, saying what was found in the reader’s terms. For example, say that the job has stopped running, not how you detected it.
• next_step: up to 300 characters, saying what a person should do. It stays empty when nothing needs doing.

The verdict is a word you choose. On the agent form, the Verdicts it can give list holds up to 10 words. Each is one word (letters, digits, dashes), listed once, and never no_op, because nothing to report is always possible. Each word comes with a note on when to give it. With a list in place, the tool accepts only those words, a finding must carry one, and an off-list or missing verdict goes back to the agent to fix within the same run.

The presets show how this maps to a stopped cron job:

• investigate is for one entity that looks like a problem and needs a closer look. A cron job that missed its schedule fits here.
• rca is for when you found what is failing and why. Use it once the agent has traced the cause of the stopped job.
• resolved is for when it was noise, or it is already over, for example a job that was late but has since run.

For a run that does not carry a headline or next step, 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:”. Agents are told to give both anyway.

Set it up

1. Open the agent that watches your cron jobs and find the How it ends section of the agent form.
2. Tick Ends every run with a conclusion (concludes: true). The agent then always gets the conclude tool, whatever its prompt says, and Tares adds the instructions for it. Your prompt only needs to describe the job. If a run stops without concluding, Tares asks once more with conclude as the only choice.
3. Under Verdicts it can give, add a word for the stopped-job case and say when to give it. The Triage (on a schedule) preset already comes with investigate. Root cause after a handoff comes with rca and resolved.
4. Keep the prompt about the job itself, for example checking whether the cron job ran on schedule. Leave the wording of the conclusion to Tares.
5. Check the agent’s summary, which shows how it ends.

The same setup as 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: |
      ...

When the run finds the stopped cron job, a good record looks like this:

• Verdict: investigate while the cause is unknown, or rca once you know what is failing and why.
• Headline: one line under 100 characters that names the job and says it stopped.
• Next step: a short instruction a person can act on, up to 300 characters.

For general background on running an AI SRE with Tares, start at AI SRE.

What you get

• A finding with a verdict, so a stopped cron job is clearly marked as something to look at, a diagnosed cause, or resolved.
• A headline of up to 100 characters and a next step of up to 300 characters. The project’s Overview shows the headline and next step on each card.
• No finding at all when the job is fine, because no_op records nothing and keeps quiet runs out of your results.
• Predictable verdicts. With a list on the agent, an off-list word or a missing verdict goes back to the agent to fix within the same run.
• Handoffs that work. The agent’s handoffs key off these verdict words, so investigate from a triage agent can pass the job on to an agent that finds the root cause.

When something else fits

• If you leave Ends every run with a conclusion off, the agent’s last message is its finding, with no verdict. That suits a free-form note, but nothing can key off a verdict.
• If you do not set a verdict list, a finding can carry any verdict or none. Set a list when you want consistent words across runs and handoffs.

Related