Proof claim
The work entered through a governed PR spine with scoped intent, acceptance criteria, evidence, verification truth, and review notes.
Starter Front Door proof
The first governed PR is the proof object.
Readiness points to the work. The governed PR shows whether the work is reviewable.
This page shows the reviewer object: one Starter Front Door PR with scoped intent, acceptance criteria, evidence, verification truth, limits, and human-controlled review.
Controlled proof: SDF makes work reviewable; it does not approve, merge, repair, deploy, or enforce automatically.
Core proof object
A governed PR turns agent output into a reviewable decision surface.
See what changes when agentic work lands as a governed PR instead of an unstructured AI output.
Evidence kept close
The PR body links to committed evidence instead of asking the buyer to trust an AI summary.
Human boundary
Review and merge stay with people. automatic_execution_permitted: false remains part of the operating boundary.
Useful limit
This supports reviewer confidence on bounded work. It does not certify correctness, enforce production governance, or prove savings.
Where proof fits
The governed PR is the bridge from assessment to operating model.
The page does not replace the assessment journey. It shows the proof object that should follow when readiness and scope support a safe first governed change.
Readiness check
Check fit and obvious blockers before asking for paid work.
Paid assessment
Use agreed scope and evidence to choose the safest proof.
Starter Front Door PR
Run one bounded change so reviewers can inspect intent, evidence, checks, risks, and limits.
Customer Front Door
Turn a useful proof into a customer-specific operating model.
Demo walkthrough
Watch the governed PR proof walkthrough
A short Loom compares the same Codex Cloud instruction with and without SDF: one ordinary AI PR and one governed PR with reviewer focus, run context, playbooks, verification, evidence links, and explicit limits.
Video not loading? Watch the walkthrough on Loom.
The recording supports the proof story. The buyer path still starts with readiness and assessment before any suitable proof or operating model.
What the walkthrough proves
From AI output to governed PR evidence.
The walkthrough compares the same Codex Cloud instruction with and without SDF. This is proof of reviewer confidence on bounded work, not a claim that a customer repo is enforced.
The point is a smaller reviewer decision: what changed, what evidence exists, what passed, what was not run, and what remains outside the claim.
Ordinary AI PRs make reviewers reconstruct the story
A normal AI PR may show description, tests, and CI while leaving intent, standards, evidence, risk, and non-goals scattered.
The governed PR becomes the decision surface
SDF puts review focus, run context, prompt/run log, acceptance criteria, and evidence links into the place the reviewer already uses: the PR.
Review focus stays close to verification
The record shows which checks ran, what passed, what did not run, and where evidence lives.
Limits stay visible
The PR can be useful without pretending to approve, merge, repair, deploy, enforce, certify correctness, or produce billing-grade cost and savings data.
Next step
Start with readiness, not an install.
The proof shows where the path can lead. The current public path starts with a readiness check and, where useful, a paid assessment.
That assessment decides whether a Starter Front Door governed PR proof is the right next step.
Reusable pattern
The useful proof is the reviewer trail.
The task, standards, checks, evidence, and limits stay visible before a human decides what to do with the PR.
Cloud-agent comparison
The walkthrough compares the same cloud-agent request with and without SDF so the reviewer difference is easy to see.
Receiver-safe Front Door pattern
The governed PR spine gives agents instructions, standards, review boundaries, and evidence expectations before work starts.
Controlled receiver proofs
The pattern has been rehearsed in controlled Rails and React/TypeScript-shaped receivers with local playbooks layered onto portable SDF guidance.
What this does not claim
Governed PR proof is not customer enforcement.
This is a controlled governed PR proof, not a claim of hosted enforcement or production/customer governance. It shows the PR review trail, evidence shape, playbook application, verification truth, and human-controlled review for bounded work. It does not claim code correctness, guaranteed correctness, automatic approval, automatic merge, automatic repair, automatic deployment, security certification, billing-grade cost reporting, measured savings, or universal stack coverage.