Appearance
Shadow Harness Architecture
Components
| Component | Responsibility | Boundary |
|---|---|---|
| Capture adapter | Convert live traces into a stable envelope | Redacts before persistence |
| Case registry | Store immutable task and dataset versions | Content-addressed identity |
| Replay planner | Select cases and candidate configurations | No execution privileges |
| Executor | Run bounded candidate workloads | Isolated credentials and quotas |
| Grader | Produce versioned outcomes and rationales | Cannot mutate task records |
| Comparator | Pair baseline and candidate results | Operates on immutable runs |
| Review surface | Slice results and inspect examples | Read-only release evidence |
Data contracts
A case describes the task and expected evaluation method. A configuration fingerprints every behavior-affecting setting. A run binds cases to one configuration. A result contains trace references, output, operational measures, and grader outcomes.
Safety model
Replay is deny-by-default for side effects. Credentials are scoped by executor and environment. Sensitive fields are classified at capture. Retention applies independently to payloads and metadata. All access to raw content is auditable.
Release gate
A candidate is eligible only when required slices meet quality and reliability thresholds, no critical safety regression appears, cost remains inside budget, and failure examples have been reviewed. The gate creates evidence; it does not replace engineering judgment.