CI-native quality governance
Adopt Valinor and your coding agents inherit a senior-engineering doctrine — shipping better work before a line ever reaches CI. Then gates, a strict review rubric, and a whole-repo audit hold every change to the bar.
valinor audit
cambercreative/valinor
Valinor's own live grade (illustrative trend) — see the full audit on the workspace.
Consumer #0 of its own gates
Every gate on this page runs on the public repo that ships it — the grade card above is Valinor's own live audit, not a mockup.
- Full gate suite · every PR
- 5/5 review to merge
- Public + continuously audited
The problem
The bar
Most teams keep quality in a contributing guide and hope the agent read it. Nothing writes the bar down where CI can hold it — so it slips, one change at a time, until the codebase no longer matches the standard anyone thinks they set.
- The bar lives in someone’s head, not where the code runs.
- It drifts a little on every change — silently.
- Lazy or incomplete agent work merges before anyone notices.
Before the first gate runs
Your agents inherit
The doctrine lives in the AGENTS.md and CLAUDE.md your agents already read — a portable, version-stamped methodology that turns loose prompting into senior-engineering practice. It takes effect on the first task, long before a line of code reaches CI.
A senior-engineering lifecycle
Research before planning, plan before executing, then build and document. Your agents stop diving straight at the keyboard and start working the way a staff engineer does.
Conduct, not vibes
DRY, YAGNI, verify-don’t-assume, leave no debt behind. The same hard-and-fast conduct a good engineer internalizes, stated plainly and inherited on day one.
A real Definition of DONE
Functional, secure, tested, documented for every audience, no debt left behind. Your agents aim at the finished bar, not the first thing that compiles.
Then CI holds the line
Declare the bar.
Discipline alone is not enough — it has to be enforced. So every change runs the gate that cannot be argued with, the review that reads like a senior engineer, and the grade that moves over time, composed across three planes so nothing slips between them.
Gates that cannot be argued with
Doctrine, claims, dependency health, release integrity, branch protection, all checked in CI on every change, fail-closed. A gate either holds or it does not. No opinion, no flake, no negotiation.
Review that reads like a senior engineer
Where a rule cannot be made deterministic, an AI review rubric suite holds every change to a strict confidence bar before it can merge. The judgment layer the gates cannot replace, applied to every diff.
A grade that moves over time
A whole-repo audit grades the codebase continuously and plots the trend. The score is not a vanity badge. It is the product: the line that tells you whether the bar is holding as your agents ship.
Adopt it in one command
No service to wire up, no dashboard to babysit. The contract lives in your repo and runs in the CI you already have.
- 01
Scaffold the contract
One command writes the contract into your repo: a claims registry, the review rubrics, branch protection, and the audit config. The bar is yours to set; the wiring is written for you.
npx … init . - 02
CI holds the line
On every pull request the gates run deterministically and fail-closed, and the review rubrics judge what no gate can. Nothing merges below the bar, no matter who, or what, wrote the change.
every PR · fail-closed - 03
Watch the grade hold
A whole-repo audit regrades the codebase on a cadence and plots the trend on your audit page. You see, in one line, whether the bar is holding as your agents ship.
graded each run
We run it on ourselves first
Hold your repo to the bar
Every gate on this page runs on the repo that ships it. Point it at yours and the same discipline runs in your CI, on any stack, with no service to host and no secrets to manage.