Step 2 · One small roster·Proof GatesPT-BR
Evidence-first model routing · Visual course

Proof Gates

A configured role is not a result. Cross the real boundary, keep the receipts, and name the boundary that remains blocked.

Read first (primary source)
Your Machine Keeps the Most Honest Record of You

The evidence workflow prefers logs, timestamps, and diffs over self-description. This lesson applies that rule to model routing.

By the end, you can

  • separate configuration proof from runtime proof
  • recognize fail-closed behavior
  • report a blocked boundary without overclaiming
Start simple; open the technical layer only when useful.
1

The big idea


Every claim needs a matching receipt. A unit test proves the mapping; a real command proves startup behavior; a container run would prove the sandbox.

Think of airport gates. A boarding pass gets you to a gate, not to your destination. The analogy breaks because software gates can be replayed deterministically.

In the code

The Factory preflight requires a gateway root, gateway token, and GitHub token before constructing the roster. The full workspace baseline passed. Docker remains explicitly unproved because Colima never became SSH-ready.

2

In one picture


configurationfail closed real commandreal path workspace tests1,865 passed Dockerunproved A blocked boundary stays visible; it is never converted into a success claim.
Green evidence does not erase the dashed Docker boundary.
3

In the code


packages/factory/src/FactoryRuntime.ts
const parsed = factoryEnvironmentSchema.safeParse(env);
if (!parsed.success) {
  const field = parsed.error.issues[0]?.path[0];
  return failure(
    `Missing required Factory environment variable: ${String(field ?? "unknown")}`,
  );
}

How to find or run this

FACTORY_CLIPROXY_URL= FACTORY_CLIPROXY_TOKEN= GH_TOKEN= pnpm exec tsx .factory/run.ts exits before any container or model call.

4

Try it


Toggle the two receipts. The simulator allows a ready claim only when both are present.

Boundary claim simulator

Ready to claim the complete boundary.

Both receipts are present.

Blocked boundary remains.

Report what passed and name what did not.

Flip the three cards, commit a prediction, then take the check.

1,865 tests passed. Is Docker proven?
No. Tests and Docker are different boundaries.
What does fail closed mean?
Stop before a model call or write when required wiring is absent.
What is a receipt?
A command, result, log, or artifact tied to the claim.
The config test passes but the daemon is unavailable. What is the honest status?
Configuration and tests are proven; the container runtime boundary is blocked.
Which report is evidence-first?
Evidence-first reporting preserves both the green gates and the unresolved boundary.
Next: apply the same honesty rule to a local DeepSeek route.

Sources and receipts