Run your first failure check

An audit turns a set of sessions into ranked, evidence-backed failure findings. Start with one narrow failure question.

Dashboard
CLI

Open Audits

In the Cloud sidebar, go to Analyze → Audits. Select new audit.

Set up the audit

Enter a name and a direct failure goal, such as “find production sessions that retry the same failed payment call without changing input or escalating.” Select the agent, environment, time window, and schedule. Add a short brief and reference URLs when the auditor needs your workflow rules.

Create and wait for the run

Select create audit. The first run is queued immediately. Open the audit card to watch it move from queued to running to complete.

After the run completes, open a finding and review its session evidence. You can then acknowledge, assign, dismiss, mute, or resolve it.

A completed audit finding showing its failure description, recommended action, supporting evidence, and affected sessions.

fp audits create payment-retry-failures \
  --description "Find production sessions that retry a failed payment call without changing input or escalating" \
  --scope '{"environments":["production"]}' \
  --text "Escalate after two unchanged payment failures"

fp audits list
fp audits runs payment-retry-failures
fp audits findings --audit payment-retry-failures
fp audits finding <finding-id>

Add reference context, then refresh it when the source changes:

fp audits context-set payment-retry-failures --url https://example.com/payment-runbook
fp audits context-refresh payment-retry-failures
fp audits run payment-retry-failures

Use fp audits ack <finding-id>, fp audits assign <finding-id> --to <email>, fp audits dismiss <finding-id>, fp audits mute <finding-id>, fp audits resolve <finding-id>, or fp audits reopen <finding-id> to move a finding through triage.

Next, prevent your first failure with a policy for a confirmed repeatable action.