Intent
What the change was supposed to do.
Change Record
Give every AI pull request a durable Change Record.
Agents say a change is done. Abizor checks whether the code, evidence, tests, and PR state support that claim, then projects the run into a Change Record a reviewer can understand.
Canonical truthevent log
Read modelChange Record
OutputJSON + Markdown
Verifierverify-proof
This page is for
Reviewers who need to understand the record left after a run.
Use this page when the question is artifact-specific: what the Change Record JSON and Markdown preserve after Abizor runs, and how that record remains a projection over the event log.
What it contains
The state behind the pull request, not another agent summary.
Every AI PR gets an Abizor Change Record: intent, scope, facts, receipts, readiness, and delivery boundary.
The Change Record is deterministic and projection-only. If the event log changes, the Change Record can be regenerated and checked again.
Why it matters
AI speed breaks down when work state is only in chat.
Scope
Which files and public surfaces changed.
Facts
Observed repository, pull request, CI, and authority state.
Receipts
What Abizor observed at each step.
Readiness
Whether fresh facts and receipts support the next step.
Delivery boundary
Whether the next delivery step is ready, blocked, or needs input.
Projection boundary
The Change Record is not a new source of truth.
Canonical truth remains the Abizor event log. The Change Record is a read model over that log, and the GitHub Check, PR comment, summary, proof report, and public proof page are projections or evidence.
abizor verify-proof recomputes hashes, regenerates the
Change Record from the event log, and returns a short verified or
blocked result.
See a real Change Record artifact proof.
Abizor has exported Change Record JSON and Markdown from real GitHub Actions runs.
Read Change Record proof Run the CI proof