Sandbox eval planeWave 1 governed shell
WorkspaceAweb / command-private

Sandbox scorecards stay durable before they influence production promotion.

This view reads the Stage 7 eval ledger for agent, prompt, provider, and workflow scorecards produced by governed E2B sandbox batches.

Private route

This shell inherits authenticated page-route posture and is not exposed as a public marketing surface.

Bounded authority

Runner registration and connector mutation stay blocked; signed Mission approval issuing, sandbox eval summaries, cost visibility, and operator sandbox interrupts are live.

Manual API gate

Published /api/v2/os9 fleets, runners, approvals, and mission contracts use route-level auth and explicit OpenAPI registration; broader OS9 APIs remain closed.

Eval runs0

No summary loaded.

Pass raten/a

0 passed, 0 failed cases.

Mean scoren/a

Weighted by case count across stored eval runs.

Eval cost$0.00

Actual cost when present, otherwise estimated sandbox cost.

Durable eval ledger

Runs are written through the API Warehouse sandbox eval store with capped stdout and stderr previews. Full execution evidence stays linked through sandbox sessions.

ready
Authenticated operator session required before reading sandbox eval state.