AI Agent Handoff Auditor audits AI-assisted engineering work before it is handed to another developer or AI agent. It reconstructs what was actually completed, maps changed files and implementation decisions, separates verified evidence from completion claims, audits available tests and verification results, identifies missing context, and exposes unresolved risks or blockers.

Instead of simply summarizing an AI coding session, it produces an evidence-backed Handoff Audit with a clear READY, READY WITH LIMITATIONS, NOT READY, or BLOCKED verdict.

The workflow is designed for teams using AI coding agents where work can appear complete while important decisions, tests, assumptions, configuration requirements, or unresolved issues remain undocumented.

It produces a concrete continuation package containing the task scope, completed work, change map, verification evidence, assumptions, risk register, limitations, and prioritized next-owner actions.

It is evidence-first, privacy-conscious, and non-destructive: it does not expose secrets, invent verification, or modify the repository merely to perform an audit.