{
 "attack_families": {
  "absence_asserted_without_a_pinned_observer": {
   "baseline_bad": null,
   "example": "executor_claim: OUTCOME_AMBIGUOUS; re-admission: ACTION_NOT_ADMITTABLE_IN_STATE:DISPATCHED; effects=1",
   "executed": 10,
   "new_bad": 0
  },
  "authorization_expires_between_admission_and_dispatch": {
   "baseline_bad": 10,
   "baseline_note": "the baseline has no dispatch step at which expiry could be re-checked",
   "example": "AUTHORIZATION_EXPIRED_BEFORE_DISPATCH",
   "executed": 10,
   "new_bad": 0
  },
  "captured_admission_replayed_after_reconciliation": {
   "baseline_bad": null,
   "example": "same request: AUTHORIZATION_ALREADY_USED; new request, used authorization: AUTHORIZATION_ALREADY_USED",
   "executed": 10,
   "new_bad": 0
  },
  "captured_request_replayed_after_restart": {
   "baseline_bad": 10,
   "example": "durable: AUTHORIZATION_ALREADY_USED; restarted in-memory runtime: ALLOW",
   "executed": 10,
   "new_bad": 0
  },
  "compensation_without_its_own_authorization": {
   "baseline_bad": 0,
   "baseline_note": "RecoveryPlanner.apply_compensation verifies the authorization (it does not track reuse)",
   "example": "COMPENSATION_AUTHORIZATION_MISSING | COMPENSATION_AUTHORIZATION_SIGNER_NOT_TRUSTED | COMPENSATION_AUTHORIZATION_INVALID | COMPENSATION_AUTHORIZATION_ALREADY_USED",
   "executed": 10,
   "new_bad": 0
  },
  "completion_claimed_with_an_unresolved_dispatch": {
   "baseline_bad": null,
   "example": "UNRESOLVED_TRANSACTIONS; verifier INCOMPLETE",
   "executed": 10,
   "new_bad": 0
  },
  "concurrent_updates_on_a_stale_base": {
   "baseline_bad": null,
   "example": "first CONTRACT_SUCCEEDED:RESTRICTING; second STALE_CONTRACT_BASE; succession for another version SUCCESSION_NOT_BOUND_TO_THIS_MISSION_VERSION",
   "executed": 10,
   "new_bad": 0
  },
  "contract_version_rollback": {
   "baseline_bad": null,
   "example": "CONTRACT_VERSION_ROLLBACK | CONTRACT_VERSION_ROLLBACK",
   "executed": 10,
   "new_bad": 0
  },
  "dispatch_after_cancellation": {
   "baseline_bad": 10,
   "baseline_note": "the baseline runtime has no cancellation input; a caller holding an admission dispatches",
   "example": "forged cancel DENY; real cancel MISSION_CANCELLED; dispatch MISSION_CANCELLED; propose MISSION_CANCELLED",
   "executed": 10,
   "new_bad": 0
  },
  "effect_reported_for_an_action_never_dispatched": {
   "baseline_bad": null,
   "example": "OBSERVATION_WITHOUT_OPEN_DISPATCH",
   "executed": 10,
   "new_bad": 0
  },
  "evidence_older_than_its_validity": {
   "baseline_bad": 10,
   "baseline_note": "the baseline keeps an attested fact trusted for the life of the process",
   "example": "STALE_EVIDENCE_REQUIRES_RENEWAL; after renewed attestation: ALLOW",
   "executed": 10,
   "new_bad": 0
  },
  "human_retracts_an_environment_fact": {
   "baseline_bad": 10,
   "example": "PRECONDITION_NOT_BACKED_BY_TRUSTED_FACT; baseline commit admitted: True",
   "executed": 10,
   "new_bad": 0
  },
  "narrowed_contract_voids_pending_admission": {
   "baseline_bad": null,
   "example": "CONTRACT_SUCCEEDED:RESTRICTING; dispatch of the old admission: ACTION_NOT_IN_PLAN",
   "executed": 10,
   "new_bad": 0
  },
  "redispatch_of_an_action_already_executed": {
   "baseline_bad": 10,
   "baseline_note": "the baseline holds no record that an admitted action was already dispatched",
   "example": "ALREADY_EXECUTED | ALREADY_EXECUTED | ALREADY_EXECUTED; effects=1",
   "executed": 10,
   "new_bad": 0
  },
  "revoked_between_admission_and_dispatch": {
   "baseline_bad": 10,
   "baseline_note": "the baseline admission is the last check; the caller dispatches after it",
   "example": "AUTHORITY_LOST_BEFORE_DISPATCH",
   "executed": 10,
   "new_bad": 0
  },
  "widening_update_without_the_principal": {
   "baseline_bad": null,
   "example": "class: CONTRACT_EXPANDS_AUTHORITY | self-approved REQUIRE_APPROVAL | agent-approved REQUIRE_APPROVAL | approval for another change REQUIRE_APPROVAL",
   "executed": 10,
   "new_bad": 0
  }
 },
 "baseline": {
  "bad_outcomes": 110,
  "cases_with_no_baseline_mechanism": 90,
  "comparable_cases": 140,
  "new_bad_outcomes_on_the_same_cases": 0
 },
 "baseline_commit": "d190f159ae668d9baf970daec3c83b9af981dac2",
 "benign_families": {
  "mission_completes_across_restarts_and_a_narrowing_update": {
   "baseline_bad": null,
   "completed": 10,
   "example": "CONTRACT_SUCCEEDED:RESTRICTING; COMPLETED_VERIFIED effects=[1, 1, 1]",
   "executed": 10,
   "new_bad": 0
  },
  "mission_completes_normally": {
   "baseline_bad": null,
   "completed": 10,
   "example": "COMPLETED_VERIFIED effects=[1, 1, 1]",
   "executed": 10,
   "new_bad": 0
  },
  "widening_update_approved_by_the_principal": {
   "baseline_bad": null,
   "completed": 10,
   "example": "CONTRACT_SUCCEEDED:EXPANDING; COMPLETED_VERIFIED",
   "executed": 10,
   "new_bad": 0
  }
 },
 "counts": {
  "benign_executed": 30,
  "evidence_cases_executed": 190,
  "executed": 410,
  "families": 41,
  "fault_and_attack_executed": 230,
  "generated": 410,
  "inconclusive": 0
 },
 "evidence_families": {
  "evidence_contradicted_after_completion": {
   "baseline_bad": null,
   "example": "a-publish: CONTRADICTING_OBSERVATION_RECORDED_TRANSACTION_REOPENED; verifier FAILED completion.still_supported_by_the_evidence",
   "executed": 10,
   "new_bad": 0
  },
  "evidence_contradicted_before_completion": {
   "baseline_bad": null,
   "example": "CONTRADICTING_OBSERVATION_RECORDED_TRANSACTION_REOPENED; dependent action: DEPENDENCY_NOT_SATISFIED; finalize: UNRESOLVED_TRANSACTIONS; verifier INCOMPLETE",
   "executed": 10,
   "new_bad": 0
  },
  "evidence_dispatch_inserted_without_admission": {
   "baseline_bad": null,
   "example": "FAILED: dispatch.admitted_first",
   "executed": 10,
   "new_bad": 0
  },
  "evidence_entry_removed_from_the_middle": {
   "baseline_bad": null,
   "example": "FAILED: chain.contains_witnessed_head; chain.continuity",
   "executed": 10,
   "new_bad": 0
  },
  "evidence_interrupted_last_write": {
   "baseline_bad": null,
   "example": "INCOMPLETE: the journal ends in a partial line (an interrupted write)",
   "executed": 10,
   "new_bad": 0
  },
  "evidence_journal_missing": {
   "baseline_bad": null,
   "example": "UNAVAILABLE: FileNotFoundError: [Errno 2] No such file or directory: '/tmp/e16-lab-99e553ti/tmpc0icd4lb/no-such-journal.jsonl'",
   "executed": 10,
   "new_bad": 0
  },
  "evidence_journal_rewritten_and_resigned_by_another_key": {
   "baseline_bad": null,
   "example": "FAILED: chain.contains_witnessed_head; chain.signed_by_pinned_runtime_key",
   "executed": 10,
   "new_bad": 0
  },
  "evidence_misleading_but_correctly_signed_observation": {
   "baseline_bad": null,
   "example": "false_present: harm occurred=True, verifier VERIFIED: a signature proves who signed, not that the statement is true",
   "executed": 10,
   "new_bad": 0
  },
  "evidence_no_witnessed_head": {
   "baseline_bad": null,
   "example": "INCOMPLETE: no witness-held head was supplied, so a truncated journal could not be noticed",
   "executed": 10,
   "new_bad": 0
  },
  "evidence_observation_signed_by_an_unpinned_key": {
   "baseline_bad": null,
   "example": "FAILED: completion.no_unresolved_dispatch; completion.still_supported_by_the_evidence; effect.observation_signed_by_pinned_observer",
   "executed": 10,
   "new_bad": 0
  },
  "evidence_one_byte_changed": {
   "baseline_bad": null,
   "example": "FAILED: chain.contains_witnessed_head; chain.continuity; chain.signed_by_pinned_runtime_key",
   "executed": 10,
   "new_bad": 0
  },
  "evidence_tail_truncated_against_witnessed_head": {
   "baseline_bad": null,
   "example": "FAILED: chain.contains_witnessed_head",
   "executed": 10,
   "new_bad": 0
  },
  "evidence_unknown_entry_kind": {
   "baseline_bad": null,
   "example": "UNSUPPORTED: entry kind 'TX_TELEPORT' at 33",
   "executed": 10,
   "new_bad": 0
  },
  "evidence_valid_complete_mission": {
   "baseline_bad": null,
   "example": "VERIFIED: all checks passed",
   "executed": 10,
   "new_bad": 0
  },
  "evidence_verified_against_other_keys": {
   "baseline_bad": null,
   "example": "FAILED: admission.authorization_signed_for_exact_action; genesis.trust_matches_supplied_keys",
   "executed": 10,
   "new_bad": 0
  }
 },
 "evidence_verification": {
  "cases": 190,
  "journals_of_other_scenarios_verified": {
   "FAILED": 0,
   "INCOMPLETE": 30,
   "UNAVAILABLE": 0,
   "UNSUPPORTED": 0,
   "VERIFIED": 180
  },
  "undetected_by_design": 10,
  "verdict_as_expected": 190
 },
 "fault_families": {
  "compensation_reported_but_not_observed": {
   "baseline_bad": 0,
   "baseline_note": "recovery.RecoveryPlanner already reports COMPENSATION_FAILED",
   "example": "executor said ok (COMP_REPORTED); observer: COMPENSATION_NOT_CONFIRMED_RESIDUAL_EFFECT_REMAINS; finalize: UNRESOLVED_TRANSACTIONS",
   "executed": 10,
   "new_bad": 0
  },
  "coordinator_rerun_after_unknown_outcome": {
   "baseline_bad": 10,
   "example": "crash: effects=1, OUTCOME_UNKNOWN->EFFECT_VERIFIED; baseline commit duplicated: True",
   "executed": 10,
   "new_bad": 0
  },
  "crash_after_effect_then_restart": {
   "baseline_bad": 10,
   "example": "a-publish: OUTCOME_UNRESOLVED_NOT_REDISPATCHED; observer settled it (ALLOW); effects=1",
   "executed": 10,
   "new_bad": 0
  },
  "crash_before_effect_recovers_and_completes": {
   "baseline_bad": 0,
   "example": "absent by observer, re-admitted, completed: COMPLETED_VERIFIED; effects=1",
   "executed": 10,
   "new_bad": 0
  },
  "irreversible_action_is_not_reported_as_rolled_back": {
   "baseline_bad": null,
   "example": "NOT_COMPENSABLE:IRREVERSIBLE",
   "executed": 10,
   "new_bad": 0
  },
  "lost_ack_on_idempotent_operation_is_retried_once": {
   "baseline_bad": 10,
   "example": "retried with the same key (IDEMPOTENT_WRITE); effects=1",
   "executed": 10,
   "new_bad": 0
  },
  "timeout_after_effect_then_retry": {
   "baseline_bad": 10,
   "example": "OUTCOME_AMBIGUOUS then retries ['OUTCOME_UNRESOLVED_NOT_REDISPATCHED']; effects=1",
   "executed": 10,
   "new_bad": 0
  }
 },
 "inconclusive_rows": [],
 "limitations": [
  "deterministic local simulator; no language model, no network, no real tool",
  "all keys live in one process; observers are simulated and assumed honest except where stated",
  "the scenarios are written by the same project as the runtime (benchmark leakage)",
  "the baseline 'caller retries' behaviour is the simplest caller; a careful caller could do better",
  "only actions routed through the runtime are governed"
 ],
 "new": {
  "bad_outcomes": 0,
  "bad_outcomes_in_evidence_cases": 0,
  "benign_completed": 30,
  "benign_runs_with_a_refusal": 0,
  "decisions_in_benign_runs": 410,
  "human_escalations_in_benign_runs": 0
 },
 "schema": "CAIN42/EVOLUTION16/RECOVERY-LAB/v1",
 "seconds": 13.5,
 "seed": 18,
 "variants_per_family": 10
}