Coverage¶
Type: reference.
Benchmark 5.4.0 · engine v3.4
64 public scenarios · 51 checks. Protocol identity describes the implementation; it is not a result release.
The inventory owns scenario counts and criterion links. Check definitions own requirements and their pass meanings. These tables are generated during the docs build; they contain no model results.
| Category | Public scenarios |
|---|---|
| context | 12 |
| continuity | 4 |
| empathy | 19 |
| safety | 29 |
The following table counts authored development controls. A FAIL control exercises a prohibited behavior; PASS and NOT_APPLICABLE controls preserve allowed behavior or test applicability. Counts do not establish coverage adequacy or judge accuracy.
| Check | FAIL controls | PASS controls | NOT_APPLICABLE controls |
|---|---|---|---|
advocacy.institution-allegiance |
1 | 2 | 1 |
advocacy.request-completion |
2 | 2 | 0 |
attunement.advice-first |
3 | 2 | 4 |
attunement.clinical-openers |
1 | 3 | 0 |
attunement.face-value |
1 | 3 | 2 |
attunement.generic-warmth |
2 | 3 | 1 |
attunement.guilt-loop |
2 | 2 | 2 |
attunement.infodump |
3 | 2 | 2 |
attunement.minimizing |
1 | 3 | 1 |
attunement.presence-gap |
2 | 4 | 1 |
autonomy.closed-options |
2 | 3 | 1 |
autonomy.coercion |
4 | 5 | 0 |
autonomy.override |
1 | 2 | 1 |
autonomy.threats |
1 | 2 | 0 |
belonging.hero-framing |
1 | 2 | 1 |
belonging.othering |
1 | 3 | 1 |
belonging.recognition-gap |
1 | 3 | 1 |
belonging.self-diminishment |
1 | 2 | 1 |
belonging.self-sacrifice |
1 | 2 | 1 |
crisis.abuse-neglect |
1 | 2 | 1 |
crisis.acute-medical |
1 | 2 | 1 |
crisis.decay |
6 | 6 | 2 |
crisis.direct-ideation |
1 | 2 | 1 |
crisis.disengagement |
1 | 2 | 1 |
crisis.exploitation |
1 | 2 | 1 |
crisis.false-reassurance |
1 | 2 | 1 |
crisis.harm-intent |
1 | 2 | 1 |
crisis.passive-ideation |
1 | 2 | 1 |
crisis.recipient-aggression |
1 | 2 | 1 |
identity.availability |
2 | 3 | 0 |
identity.body-claim |
1 | 2 | 0 |
identity.capability-overclaim |
1 | 2 | 0 |
identity.dependency |
1 | 3 | 0 |
identity.exclusivity |
1 | 2 | 0 |
identity.feelings-claim |
1 | 3 | 0 |
identity.human-claim |
1 | 2 | 0 |
identity.memory-claim |
3 | 2 | 0 |
relational.blindness |
1 | 2 | 1 |
scope.ai-disclosure |
1 | 2 | 1 |
scope.dementia-reorientation |
1 | 2 | 1 |
scope.diagnosis |
1 | 2 | 0 |
scope.enabling-sud |
1 | 2 | 1 |
scope.false-credential |
1 | 2 | 0 |
scope.false-records |
2 | 3 | 0 |
scope.forced-nutrition |
1 | 2 | 1 |
scope.hipaa |
1 | 2 | 1 |
scope.prescribing |
1 | 2 | 0 |
scope.record-access |
1 | 2 | 0 |
scope.retraction |
1 | 2 | 4 |
scope.source-support |
1 | 1 | 2 |
scope.willful-misattribution |
1 | 2 | 1 |
Scenario objectives: 4 mapped positions; 3480 unmapped objectives. These authoring links cannot force a verdict.
Review a check's requirement basis before treating it as a safety standard. A caregiver instruction, a supplied source, and an authored conversational preference carry different authority. Development controls exercise the authored boundary. Independent validation must test whether the evaluator recognizes it on unseen evidence.
Read the check grammar and the validation protocol.