Controlled rehearsal proof
A controlled Campfire-style proof shows SDF attaching governed evidence to cloud-agent work, checking the PR reviewer surface, and catching incomplete or wrongly targeted PR publication.
The demo proof page shows controlled proof of the operating model: agent-generated work moves with evidence, the reviewer-facing PR description is checked against that evidence, and description remediation happens only with explicit permission.
Controlled proof only. No hosted enforcement, production/customer governance, automatic approval, automatic merge, automatic code repair, security certification, or guaranteed correctness is claimed.
Core proof line
Homepage proof stays buyer-friendly. This page carries the deeper evidence chain and boundaries.
Demo walkthrough
A short Loom walkthrough shows the current controlled path from assisted assessment context to completed governed PR proof.
Some browsers block embedded Loom playback. The direct link remains the reliable demo path.
Video not loading? Watch on Loom
The recording is a walkthrough of controlled rehearsal proof surfaces. The current public path starts with a private repo review, then a Front Door install and first governed change where the repo is suitable.
This is a controlled rehearsal proof walkthrough. It does not claim hosted enforcement, production/customer governance, automatic approval, automatic merge, billing-grade cost, or measured savings.
What the walkthrough proves
The walkthrough proves the full path from adoption to governed work without turning the GTM site into an execution surface.
This is what governed AI-assisted delivery looks like in practice: a bounded change, governed evidence, a reviewable PR surface, and committed evidence for drill-down.
SDF caught an incomplete PR reviewer surface, remediated the PR description from governed evidence with explicit permission, and detected a wrong-base PR publication issue. It made the delivery trail visible; it did not approve, merge, deploy, repair, or mutate application code.
Controlled Rails and React/TypeScript rehearsal receivers have completed bounded governed work with intent, acceptance criteria, risk notes, verification evidence, and handoff context attached.
The current proof runs as local SDF tooling inside controlled rehearsals. It is not hosted enforcement, automatic repo mutation, automatic approval, automatic merge, automatic deploy, or automatic repair.
GTM dogfooding and controlled receiver work leave evidence. Reviewed lessons can become safer Front Door guidance without claiming autonomous policy mutation or automatic optimisation.
Next step
The demo shows where the path can lead. The current public path starts with a repo review.
We use lightweight repo signals and shared context to decide whether SDF Front Door is a fit before any install or governed change.
Completed governed work proofs
These are controlled rehearsal proofs, not a claim of production governance, all-stack coverage, or native tooling for every receiver language.
A controlled Campfire-style receiver proof shows SDF can make cloud-agent work reviewable by attaching governed evidence, checking the PR reviewer surface, catching incomplete or wrongly targeted PR publication, and remediating PR descriptions from governed evidence with explicit permission.
A governed work item has been completed end-to-end in a controlled Rails receiver.
A governed work item has been completed end-to-end in a controlled React/TypeScript-shaped receiver, with the Ruby-based SDF factory tooling running separately from the Node app.
What this does not claim
This is controlled rehearsal proof, not a claim of production customer enforcement. It shows the governed workflow, evidence shape, and review surface working in bounded receiver repos. It does not claim code correctness, hosted enforcement, automatic approval or merge, automatic repair, guaranteed correctness, security certification, billing-grade cost reporting, measured savings, or universal stack coverage.
