CAIN-42 CAIN Studio

Evidence library · niche

Autonomy, control loops & recovery

Agents acting on their own, and how CAIN keeps them in bounds and recovers. 407 tested invariants.

Last reviewed 2026-10-01

407 of 407 held

Where this niche's rules come from

Test families in this niche

evolution (56) · loop (44) · autonomy (25) · drift (25) · learning_loop (17) · promotion_dominance (15) · self_healing (15) · self_test (13) · degradation (13) · self_governance (12) · autonomy_gradient (11) · self_improvement (10) · promotion_stage (10) · self_repair (10) · degraded_mode (10) · replan (10) · continuity (9) · autonomy_level (9) · research (7) · autopilot (6) · recovery (6) · fabric_component (4) · swe (3) · state_machine (2) · integrity (2) · profile (2) · risk (1) · scientist (1) · operation_state (1) · gap_detector (1) · plan_check_act (1) · trajectory (1) · radar (1) · swarm (1) · continuity_token (1) · canary (1) · interrupt (1) · promotion (1) · replanning (1) · rollback (1) · systemic (1) · graph (1)

Where these come from

Rules 151–300 of 407

IDRuleBundleResult
E32-SELFTEST-enforcementself-test of enforcement passes, and a failing check degrades itE32held
E32-SELFTEST-evidenceself-test of evidence passes, and a failing check degrades itE32held
E32-SELFTEST-memoryself-test of memory passes, and a failing check degrades itE32held
E32-SELFTEST-world_modelself-test of world_model passes, and a failing check degrades itE32held
E32-SELFTEST-learningself-test of learning passes, and a failing check degrades itE32held
E32-SELFTEST-evolutionself-test of evolution passes, and a failing check degrades itE32held
E32-SELFTEST-rollbackself-test of rollback passes, and a failing check degrades itE32held
E32-SELFTEST-revocationself-test of revocation passes, and a failing check degrades itE32held
E32-SELFTEST-clean_roomself-test of clean_room passes, and a failing check degrades itE32held
E32-SCI-promotethe governance scientist may not promoteE32held
E32-REPAIR-ISOLATEgoverned self-repair records ISOLATE in orderE32held
E32-REPAIR-PRESERVE_EVIDENCEgoverned self-repair records PRESERVE_EVIDENCE in orderE32held
E32-REPAIR-IDENTIFYgoverned self-repair records IDENTIFY in orderE32held
E32-REPAIR-REPAIR_CANDIDATEgoverned self-repair records REPAIR_CANDIDATE in orderE32held
E32-REPAIR-TESTgoverned self-repair records TEST in orderE32held
E32-REPAIR-ADVERSARIAL_TESTgoverned self-repair records ADVERSARIAL_TEST in orderE32held
E32-REPAIR-CLEAN_ROOM_VERIFYgoverned self-repair records CLEAN_ROOM_VERIFY in orderE32held
E32-REPAIR-CANARYgoverned self-repair records CANARY in orderE32held
E32-REPAIR-PROMOTEgoverned self-repair records PROMOTE in orderE32held
E32-REPAIR-ROLLBACK_IF_REGRESSIONgoverned self-repair records ROLLBACK_IF_REGRESSION in orderE32held
E32-LOOP-OBSERVEloop step OBSERVE holds on real governed actionsE32held
E32-LOOP-EVIDENCEloop step EVIDENCE holds on real governed actionsE32held
E32-LOOP-PREDICTloop step PREDICT holds on real governed actionsE32held
E32-LOOP-ACTloop step ACT holds on real governed actionsE32held
E32-LOOP-OUTCOMEloop step OUTCOME holds on real governed actionsE32held
E32-LOOP-COMPAREloop step COMPARE holds on real governed actionsE32held
E32-LOOP-ERROR_DRIFT_FAILUREloop step ERROR_DRIFT_FAILURE holds on real governed actionsE32held
E32-LOOP-LEARNloop step LEARN holds on real governed actionsE32held
E32-LOOP-GENERATE_CANDIDATEloop step GENERATE_CANDIDATE holds on real governed actionsE32held
E32-LOOP-SANDBOXloop step SANDBOX holds on real governed actionsE32held
E32-LOOP-ADVERSARIAL_TESTloop step ADVERSARIAL_TEST holds on real governed actionsE32held
E32-LOOP-PROOFloop step PROOF holds on real governed actionsE32held
E32-LOOP-CONFORMANCEloop step CONFORMANCE holds on real governed actionsE32held
E32-LOOP-CANARYloop step CANARY holds on real governed actionsE32held
E32-LOOP-RE_AUTHORIZEloop step RE_AUTHORIZE holds on real governed actionsE32held
E32-LOOP-PROMOTE_OR_REJECTloop step PROMOTE_OR_REJECT holds on real governed actionsE32held
E32-LOOP-OBSERVE_AGAINloop step OBSERVE_AGAIN holds on real governed actionsE32held
E32-DOM-authority_preservationa regression in authority_preservation blocks promotionE32held
E32-DOM-enforcement_coveragea regression in enforcement_coverage blocks promotionE32held
E32-DOM-evidence_integritya regression in evidence_integrity blocks promotionE32held
E32-DOM-identity_continuitya regression in identity_continuity blocks promotionE32held
E32-DOM-revocation_latency_usa regression in revocation_latency_us blocks promotionE32held
E32-DOM-failure_containmenta regression in failure_containment blocks promotionE32held
E32-DOM-prediction_calibrationa regression in prediction_calibration blocks promotionE32held
E32-DOM-policy_correctnessa regression in policy_correctness blocks promotionE32held
E32-DOM-adversarial_robustnessa regression in adversarial_robustness blocks promotionE32held
E32-DOM-recoverya regression in recovery blocks promotionE32held
E32-DOM-observabilitya regression in observability blocks promotionE32held
E32-DOM-reproducibilitya regression in reproducibility blocks promotionE32held
E32-DOM-verifier_independencea regression in verifier_independence blocks promotionE32held
E32-DOM-computational_cost_usa regression in computational_cost_us blocks promotionE32held
E32-DOM-governance_latency_usa regression in governance_latency_us blocks promotionE32held
E33-STATE-RECOVER_ADAPTno state can be skipped after RECOVER_ADAPTE33held
E33-DEGRADED-verifier_unavailableverifier_unavailable only removes routesE33held
E33-DEGRADED-evidence_unavailableevidence_unavailable only removes routesE33held
E33-DEGRADED-trust_domain_unreachabletrust_domain_unreachable only removes routesE33held
E33-DEGRADED-policy_service_unavailablepolicy_service_unavailable only removes routesE33held
E33-DEGRADED-world_state_staleworld_state_stale only removes routesE33held
E33-DEGRADED-revocation_service_unavailablerevocation_service_unavailable only removes routesE33held
E33-DEGRADED-cluster_partitioncluster_partition only removes routesE33held
E33-DEGRADED-model_unavailablemodel_unavailable only removes routesE33held
E33-DEGRADED-runtime_failureruntime_failure only removes routesE33held
E33-DEGRADED-enforcement_boundary_unavailableenforcement_boundary_unavailable only removes routesE33held
E33-GAP-recovery_gapthe gap detector knows recovery_gapE33held
E33-AUTOPILOT-expand_authoritythe autopilot may not expand_authorityE33held
E33-AUTOPILOT-disable_controlsthe autopilot may not disable_controlsE33held
E33-AUTOPILOT-promote_itselfthe autopilot may not promote_itselfE33held
E33-AUTOPILOT-alter_trust_rootsthe autopilot may not alter_trust_rootsE33held
E33-AUTOPILOT-bypass_e8the autopilot may not bypass_e8E33held
E33-AUTOPILOT-deploy_unverified_governancethe autopilot may not deploy_unverified_governanceE33held
E35-FABRIC-drift_learnerthe intelligence fabric exposes drift_learnerE35held
E35-FABRIC-evolution_enginethe intelligence fabric exposes evolution_engineE35held
E35-FABRIC-canary_enginethe intelligence fabric exposes canary_engineE35held
E35-FABRIC-rollback_enginethe intelligence fabric exposes rollback_engineE35held
E35-AUTONOMY-planningautonomy dimension planning existsE35held
E35-AUTONOMY-executionautonomy dimension execution existsE35held
E35-AUTONOMY-temporalautonomy dimension temporal existsE35held
E35-AUTONOMY-financialautonomy dimension financial existsE35held
E35-AUTONOMY-communicationautonomy dimension communication existsE35held
E35-AUTONOMY-delegationautonomy dimension delegation existsE35held
E35-AUTONOMY-capability_acquisitionautonomy dimension capability_acquisition existsE35held
E35-AUTONOMY-environmentalautonomy dimension environmental existsE35held
E35-AUTONOMY-self_modificationautonomy dimension self_modification existsE35held
E35-AUTONOMY-replicationautonomy dimension replication existsE35held
E35-AUTONOMY-organizationalautonomy dimension organizational existsE35held
E35-DECISION-ROLLBACKplan-check-act decision ROLLBACK is first-classE35held
E35-PROFILE-autonomyagentic profile dimension autonomy existsE35held
E35-PROFILE-recoverabilityagentic profile dimension recoverability existsE35held
E35-TRAJ-DRIFTINGtrajectory state DRIFTING existsE35held
E35-RADAR-self_improvementradar topic self_improvement existsE35held
E35-SWARM-evolutionswarm role evolution existsE35held
E35-DEGRADE-NORMALdegradation state NORMAL existsE35held
E35-DEGRADE-DEGRADEDdegradation state DEGRADED existsE35held
E35-DEGRADE-RESTRICTEDdegradation state RESTRICTED existsE35held
E35-DEGRADE-READ_ONLYdegradation state READ_ONLY existsE35held
E35-DEGRADE-HUMAN_REQUIREDdegradation state HUMAN_REQUIRED existsE35held
E35-DEGRADE-CONTAINMENTdegradation state CONTAINMENT existsE35held
E35-DEGRADE-OFFLINEdegradation state OFFLINE existsE35held
E35-DEGRADE-RECOVERYdegradation state RECOVERY existsE35held
E35-HEAL-detectself-healing may detectE35held
E35-HEAL-diagnoseself-healing may diagnoseE35held
E35-HEAL-proposeself-healing may proposeE35held
E35-HEAL-sandboxself-healing may sandboxE35held
E35-HEAL-testself-healing may testE35held
E35-HEAL-canaryself-healing may canaryE35held
E35-HEAL-rollbackself-healing may rollbackE35held
E35-HEAL-request_authorizationself-healing may request_authorizationE35held
E35-HEALNOT-self_authorizeself-healing may not self_authorizeE35held
E35-HEALNOT-self_expand_authorityself-healing may not self_expand_authorityE35held
E35-HEALNOT-disable_governanceself-healing may not disable_governanceE35held
E35-HEALNOT-rewrite_verifier_to_passself-healing may not rewrite_verifier_to_passE35held
E35-HEALNOT-delete_evidenceself-healing may not delete_evidenceE35held
E35-HEALNOT-hide_incidentsself-healing may not hide_incidentsE35held
E35-HEALNOT-rewrite_historyself-healing may not rewrite_historyE35held
E35-SELFGOV-agentsCAIN governs its own agentsE35held
E35-SELFGOV-policiesCAIN governs its own policiesE35held
E35-SELFGOV-pluginsCAIN governs its own pluginsE35held
E35-SELFGOV-adaptersCAIN governs its own adaptersE35held
E35-SELFGOV-modelsCAIN governs its own modelsE35held
E35-SELFGOV-researchCAIN governs its own researchE35held
E35-SELFGOV-evolutionCAIN governs its own evolutionE35held
E35-SELFGOV-evidenceCAIN governs its own evidenceE35held
E35-SELFGOV-deploymentCAIN governs its own deploymentE35held
E35-SELFGOV-transactionsCAIN governs its own transactionsE35held
E35-SELFGOV-cloudCAIN governs its own cloudE35held
E35-SELFGOV-marketplaceCAIN governs its own marketplaceE35held
E35-EVOGATE-DISCOVERYevolution gate DISCOVERY existsE35held
E35-EVOGATE-DESIGNevolution gate DESIGN existsE35held
E35-EVOGATE-THREAT_MODELevolution gate THREAT_MODEL existsE35held
E35-EVOGATE-SANDBOXevolution gate SANDBOX existsE35held
E35-EVOGATE-UNIT_TESTevolution gate UNIT_TEST existsE35held
E35-EVOGATE-ADVERSARIAL_TESTevolution gate ADVERSARIAL_TEST existsE35held
E35-EVOGATE-DIFFERENTIAL_TESTevolution gate DIFFERENTIAL_TEST existsE35held
E35-EVOGATE-CLEAN_ROOM_TESTevolution gate CLEAN_ROOM_TEST existsE35held
E35-EVOGATE-PERFORMANCE_TESTevolution gate PERFORMANCE_TEST existsE35held
E35-EVOGATE-GOVERNABILITY_TESTevolution gate GOVERNABILITY_TEST existsE35held
E35-EVOGATE-CANARYevolution gate CANARY existsE35held
E35-EVOGATE-OBSERVEevolution gate OBSERVE existsE35held
E35-EVOGATE-PROOFevolution gate PROOF existsE35held
E35-EVOGATE-HUMAN_ADMIN_APPROVALevolution gate HUMAN_ADMIN_APPROVAL existsE35held
E35-EVOGATE-DEPLOYevolution gate DEPLOY existsE35held
E35-EVOGATE-MONITORevolution gate MONITOR existsE35held
E35-EVOGATE-ROLLBACKevolution gate ROLLBACK existsE35held
E35-LOOP-OBSERVEintelligence loop step OBSERVE existsE35held
E35-LOOP-UNDERSTANDintelligence loop step UNDERSTAND existsE35held
E35-LOOP-MODELintelligence loop step MODEL existsE35held
E35-LOOP-PREDICTintelligence loop step PREDICT existsE35held
E35-LOOP-SIMULATEintelligence loop step SIMULATE existsE35held
E35-LOOP-ADVERSARIAL_TESTintelligence loop step ADVERSARIAL_TEST existsE35held
E35-LOOP-RECOMMENDintelligence loop step RECOMMEND existsE35held

1 2 3

Other niches

Consensus & distributed systems · Attacks, threats & containment · Identity, authority & delegation · Evidence, receipts & proofs · Memory, data & privacy · Prediction, world models & simulation · Transactions, markets & economics · Tools, MCP, protocols & adapters · Policy, law & governance · Trust & reputation · Supply chain, registry & lifecycle · Benchmarks, coverage & performance · Core guarantees

Try CAIN-42 on your own agents

Create a free account and every new account starts with a 7-day trial of the full platform. Or try the sandbox first, with no account at all.

Create a free account →  ·  Try the sandbox  ·  See the whole ecosystem