Use this template for human-readable evaluation reports. Attach the harness JSON report as a machine-readable artifact.
| Field | Value |
|---|---|
| Report id | |
| Profile | |
| Domain | |
| Dataset tier | synthetic onboarding / expert-authored challenge set / consent-governed pilot data |
| Harness command | |
| CAP structural conformance report | separate artifact ref |
| Gate conclusion | external_expert_evidence_required / ready_for_gate_review / blocked |
| Report owner | |
| Case set | Case count | Synthetic count | Consent/governance refs | Source hash refs |
|---|---:|---:|---|---|
| | | | | |
| Reviewer id | Role | Qualification or profile-owner basis | Independence/conflict notes |
|---|---|---|---|
| | | | |
| Criterion | Passing threshold | Required reviewers | Blocking flags |
|---|---:|---:|---|
| | | | |
| Status | Count |
|---|---:|
| pass | |
| needs_review | |
| needs_revision | |
| blocked | |
| Case id | Review count | Status | Low criteria | Blocking flags | Retest needed |
|---|---:|---|---|---|---|
| | | | | | |
| Check | Result | Evidence |
|---|---|---|
| Shared cases exclude raw transcript/audio, raw sensitive evidence, hidden reasoning, and credentials. | | |
| Non-synthetic cases include consent or governance refs. | | |
| Reviewer notes avoid raw sensitive spans. | | |
| Dataset and report retention are documented. | | |
| Finding id | Severity | Description | Owner | Status | Retest evidence |
|---|---|---|---|---|---|
| | | | | | |
State exactly one conclusion:
external_expert_evidence_required: synthetic-only run, incomplete reviewer depth, or no qualified domain expert review.blocked: a required criterion failed, a blocking flag is unresolved, or privacy review failed.ready_for_gate_review: qualified reviewers completed the plan, no blocking findings remain, and structural CAP conformance is reported separately.Sign-off:
| Reviewer or owner | Role | Date | Notes |
|---|---|---|---|
| | | | |