FIXED-SCOPE ZINVYL SERVICE

AI Workflow Failure-Mode Audit

Audit one bounded AI-enabled workflow containing no more than eight documented steps.

$295fixed published price
3 daysafter complete accepted inputs
1included revision
Describe your need and confirm scope Inspect the public proof sample
EXACT COMMERCIAL BOUNDARY

What you receive.

The published contract below controls this service. Customer-specific work begins only after signed funding verification and complete, accepted inputs.

INCLUDED SCOPE

DELIVERABLES

  • Audit one bounded AI-enabled workflow containing no more than eight documented steps.
  • Test the supplied workflow description, sanitized traces, prompts, outputs, and approval rules against a reproducible failure matrix.
  • Deliver an evidence-cited failure register, severity and containment notes, and a prioritized remediation plan; no production changes are included.

SAFE INPUTS REQUIRED

  • One sanitized workflow description with no more than eight steps.
  • Up to ten sanitized trace, prompt, output, or policy excerpts and the buyer's acceptance criteria.

ACCEPTANCE EVIDENCE

  • The audit stays inside one documented workflow of no more than eight steps.
  • Every finding maps to supplied evidence or a reproducible test.
  • Severity and containment are explicit.
  • The remediation plan is independently reviewed before delivery.
BOUNDARIES

NOT INCLUDED

  • Production access or writes, code changes, deployment, live monitoring, red-team intrusion, model training, security certification, regulated or high-impact decision review, legal advice, or guaranteed improvement.
  • More than one workflow, more than eight workflow steps, more than ten evidence excerpts, or evidence containing credentials, personal data, regulated data, or unauthorized confidential material.

EVIDENCE LIMITS

  • The audit is limited to the supplied sanitized evidence and reproducible tests; it does not establish whole-system safety, security, compliance, accuracy, or fitness for a regulated decision.
  • Severity and recommendations are human-reviewed technical judgments, not guarantees of outcome.
  • The public proof is synthetic and is not a customer result, sale, or demand claim.

Canonical capabilityzinvyl.quality.ai-workflow-failure-audit.v1