Diagnostic · 10 questions · No email required

Is your AI agent production-ready?

Ten yes/no/partial questions against the 30-point production checklist we use on every recovery audit. Score yourself in under 2 minutes. Result stays in your browser. The full recovery diagnosis covers 30 criteria over 2 weeks for $5,000.

Start the diagnosticRecovery service

0 / 10 answered

  1. 01
    Observability

    Can you see every agent decision in real time, with PII redacted?

  2. 02
    Retry logic

    Do you have exponential backoff + a dead-letter queue for failed tool calls?

  3. 03
    Drift detection

    Are you measuring output quality over time against a baseline?

  4. 04
    Prompt injection

    Are user-supplied inputs sanitized, and outputs guarded against tool-call hallucinations?

  5. 05
    Cost guardrails

    Is there a per-task cost ceiling that kills a runaway loop?

  6. 06
    Escalation

    When the agent cannot resolve a case, does it escalate to a named human within an SLA?

  7. 07
    Rollback

    Can you revert to the last-known-good prompt + tool config in under 5 minutes?

  8. 08
    Data scope

    Does the agent run with read-only or least-privilege credentials by default?

  9. 09
    Evals

    Do you run a fixed eval suite before every prompt or model change?

  10. 10
    Runbook

    Is there a written runbook a new operator can follow at 2 a.m. without paging you?

No email required · Result stays in your browser

Not production-ready?

Recovery diagnosis covers all 30 criteria, not just 10.

The full audit measures observability, retry logic, drift detection, cost variance, prompt-injection resistance, escalation paths, data security, and 23 more. Written report, 1-hour walkthrough call. $5,000.

Start a recovery diagnosisBook a free audit