Repository navigation
v3.3.0 — Cybernetic Governance Engine
Security release. It fixes two published advisories:
| Advisory | Severity | Summary | Credit |
|---|---|---|---|
| GHSA-f6xq-23f7-9grm | 5.3 Medium | Revocation during human review not re-checked after approval (FRIA tier); trade policy limits bypassed by execute_trade_bounded and BTC case variants |
Reported by Dominik Blain (@dom-omg), Ironproof |
| GHSA-w39v-5v7h-fxg9 | 5.3 Medium | Model-controlled trader_role raises OPA fiscal limits on the committing trade path |
Found during maintainer triage |
Thank you to Dominik Blain for a careful report that came with runnable tests for each finding and Z3 counterexamples, and to Daniel Fairbanks (@danofairanks) for the reports behind #511, #512, #513, #514, #516 and #519.
Upgrading: this release has breaking changes. Each is marked Breaking below; docs/BREAKING_CHANGES_v3.md covers the GHSA-f6xq-23f7-9grm changes (GH-1 to GH-7).
Release Summary: Security release. Fixes GHSA-f6xq-23f7-9grm and GHSA-w39v-5v7h-fxg9 (both CVSS 5.3 Medium). A trade now executes at most once per
transaction_id, an MCP tool call whose outcome is unknown is no longer resent, and single-use Redis writes and evidence-stream appends are retry-safe. The advisor binds the end-user principal to a verified OIDC identity, the evidence custodian recovers a lost custody cursor from WORM attestations, and skill certification receipts bind the stages and evidence they report (cage.skill.cer.v2). Entries marked Breaking change behaviour or interfaces; see alsoBREAKING_CHANGES_v3.md.
Fixed
- Approved requests re-run the FRIA tier and answer only the holds the reviewer saw, and the trade policy holds every trade action to its limits (GHSA-f6xq-23f7-9grm, CVSS 5.3 Medium) (
fix(security)!— reported by Dominik Blain (@dom-omg), Ironproof,src/gateway/governance/governor/pipeline.py,src/gateway/governance/governor/governor.py,src/gateway/governance/jurisdiction/eu_ai_act/fria_tier.py,src/gateway/governance/governor/stages/opa.py,src/cage_finance/opa/trade_governance.rego). Thank you to Dominik Blain for a careful report, which came with runnable tests for each finding and Z3 counterexamples. Three findings. (1) POST_HITL never re-ran theEU_ECBfriatier, so a determination the normative provider revoked while a human reviewed a held request (or a provider that became unavailable, or an artefact that expired meanwhile) was never observed, and the approved request was sealed. A read-only tier now declaresrevalidated_after_approval(GovernanceTier;Truefor mutating tiers,Falseby default for read-only ones, copied byDomainTierStage), andstage_runs_under(profile, *, name, mutating, revalidated)runs every such tier under POST_HITL, selected by structure, never by name.JurisdictionContributionrefuses a tier that does not declare it;FriaTierdoes. Under POST_HITL a re-run tier's HITL finding is answered only if the approval was given against it:DeferToken.hitl_holdsreads the(tier, code)holds of the parked decision,enforce_approved_governance()passes them asPostHitlApproval.approved_holds, andrun_pipeline()drops exactly those (answered_by_approval()). A hold the reviewer was never shown is refused[APPROVAL_CONTEXT_DRIFT](unapproved_holdsin the refusal meta) and a HARD finding denies, both spending the approval; aRELIANCE_INELIGIBLEfinding defers and keeps it. Re-running the tier without answering the approved hold would have blocked every legitimate approval of aFRIA_EXTERNAL_HOLD; answering every hold would have sealed one the reviewer never saw. OPA'sMANUAL_REVIEWis answered the same way:OpaStageused to answer its own review under POST_HITL, so an approval given for another hold (a FRIA hold, for example) also satisfied an OPA review nobody was shown. It now reportsgovernance/OPA_MANUAL_REVIEWunder every profile, and only an approval given against that hold answers it.proof/model.pyreplacesjurisdiction_keeps_post_hitl_setwith claim 10post_hitl_rechecks_jurisdictionand adds claim 14approval_answers_only_its_holds; state counts are unchanged (38 / 19 / 35, EU 42). (2)trade_governance.regoapplied its RBAC ladder only toexecute_trade:execute_trade_boundedfell through to the risk-profile rules, which allowed it for any role, amount and asset, and the BTC restriction comparedcurrencyexactly, so"btc","Btc"and"BTC "passed it. Every action intrade_actions(execute_trade,execute_trade_bounded) is now held to the RBAC limits, quota rules and asset restrictions and excluded from the risk-profile rules, so a bounded trade is decided exactly likeexecute_trade, including one no risk-profile rule matched (norisk_profile, or a senior'sspeculative), which fell to the default DENY before;risk_profileis caller-supplied, so that DENY was never a control (the same caller could sendmoderateand be allowed any amount or asset). A trade'scurrencymust name a tradable currency (tradable_currencies, an allowlist:USD, the only currency the reference broker settles), so an alternate code for a restricted asset (XBT) or a look-alike spelling is never admitted, andrestricted_assets(BTC,XBT) are refused as thesymboltoo; both values are compared trimmed and upper-cased. Such a trade is an explicit DENY, and_rbac_denycarries the same condition, so the authorization-claim and token-quotaMANUAL_REVIEWrules no longer route it to a human. The advisor's trade params are deliberately not normalised: approval redemption bindssymbol,currencyandsideby exact match (_approval_covers_tradeintool_provider.py), so the policy is where spellings are normalised. (3) Thevalidate_fria()payload named no caller and no correlation id, so provider_01 could not mint aConsequenceTokenand refused every admission.FriaTiernow sends the workload identity the gateway verified for the request (current_client_identity()) asactor_id/operator_urn(""when there is none, never a request value), the active trace id ascorrelation_id, and athread_idfalling back to that trace id. Tests:tests/governor/test_jurisdiction_wiring.py(the advisory's revoke-during-review scenario, provider outage and artefact expiry at revalidation, unseen-hold drift, approved holds still sealing, no barrier headroom committed on revocation, payload identity),tests/test_formal_profile_parity.py(runtime parity with claims 10 and 14, OPA's hold included),tests/governor/test_post_hitl_approval_spend.py(an approved OPA review seals; one the reviewer never saw is drift),tests/governor/stages/test_opa_stage.py, golden scenario 09 (tests/governor/golden/fixtures.py) andcompliance/postures/us_fed/opa/constraints_test.rego(23 new cases, 19 of which fail against the previous policy). OSCAL: theeu-ai-act-art27statements insystem-security-plan-eu-ecb.yaml, AC-3 insystem-security-plan.yamland ISO 42001 A.8.4 incomponent-definition.yaml. Breaking:stage_runs_under()requires therevalidatedkeyword; a jurisdiction tier that does not declarerevalidated_after_approval = Trueis refused; a POST_HITL run whose re-run tier reports a HITL hold its approval was not given against is refused instead of sealed; so is one whose OPA verdict isMANUAL_REVIEWunless its approval was given against that review (_violations_for()drops itsprofilekeyword); OPA now denies a bounded trade above the role's limit (or holds it for review in the review band), and any trade whosecurrencyis notUSDup to letter case and surrounding spaces (missing, not a string, empty,EUR,XBT, any spelling ofBTC) or whosesymbolisBTCorXBTin any spelling, even where an authorization claim or a low token quota used to route it toMANUAL_REVIEW. - GHSA-f6xq-23f7-9grm follow-ups: the order model validates
currency, generated denylists are normalised, and the provider_01 governor path is documented (fix(security)!— from the review of the report by Dominik Blain (@dom-omg), Ironproof,src/cage_finance/models/trade_order.py,src/gateway/governance/stpa_compiler.py).TradeOrdertrims and upper-casescurrencyand refuses anything but three ASCII letters, so a look-alike spelling fails when the order is built, not only in the policy; which currencies are tradable stays the policy's decision (tests/cage_finance/test_trade_order_currency.py). RBAC rules generated by the STPA compiler compared a denylisted value exactly, the same gap as the hand-written policy:_rego_denylist_guard()now emitsupper(trim_space(input.<field>)) != "<ENTRY>", the generated AGP sentence refuses the value in any letter case or with surrounding spaces, andtrade_hazards.yamldenylistsBTCandXBT(tests/test_stpa_compiler.py,tests/test_stpa_compiler_agp.py); the committedconfig/artifacts define no RBAC rules and are unchanged.config/agent_scope.yamlrecords the tradable and prohibited currencies. The provider_01 wire harness (scripts/verify_flowsignal_wire_and_hostile.py) takes the actor fromFriaTier'scaller_identityinstead of writing it into the payload. The request facts provider_01 still takes on trust (constant mandate and KYA status, permissive request defaults) are tracked as POAM-2026-113, and the provider_01 README describes the governor path: the identity in the payload, the ConsequenceToken that decision D2 leaves off the routing seal, and the emptytarget.OPA_MIGRATION_PROCESS.mdnow names the live policy file, its Rego tests and the CI job that runs them.SECURITY.mdlists the advisory and thanks the reporter. Breaking: aTradeOrderwhosecurrencyis not three letters after trimming raisesValidationError(the broker actuator reportsORDER_CONSTRUCTION_FAILED), andorder.currencyis upper-case; generated RBAC rules refuse a denylisted value in any letter case or spacing. - The advisor binds the end-user principal to a verified OIDC identity and refuses or marks unverified principals by posture (
feat(advisor), #520 —src/governed_financial_advisor/infrastructure/auth.py,src/governed_financial_advisor/server.py,src/governed_financial_advisor/geap_app.py,src/governed_financial_advisor/infrastructure/gateway_client.py,src/governed_financial_advisor/utils/context.py,src/gateway/governance/ingress/agw_adapter.py,src/gateway/governance/seams/credential_broker.py): AddedPrincipalIdentity,verify_user_oidc_token(), andresolve_request_principal()to authenticate end-user OIDC tokens (Authorization: Bearer <jwt>,X-CAGE-User-Token, orX-GEAP-User-Assertion) against out-of-bandkid-resolved trust anchors (CAGE_USER_OIDC_JWKS_JSONorCAGE_USER_OIDC_TRUST_ANCHORS_FILE) while rejecting embedded key headers (jwk,jku,x5c,x5u). A verified principal is carried end-to-end throughuser_context,principal_verified_context=True, span attributesenduser.id+cage.principal.verified=true, and outboundGatewayClient._build_headers()headersx-cage-principal,x-cage-principal-verified: true, andx-geap-user-id; a bodyuser_idthat conflicts with the authenticated token claim is refused with HTTP 403 before graph execution. Requests without a verified OIDC token are refused with HTTP 401 whenrequires_verified_principal()holds (enforcing posturesproduction/staging, whenCAGE_USER_OIDC_ISSUERis configured, or whenCAGE_REQUIRE_VERIFIED_PRINCIPAL=true, unlessCAGE_ALLOW_ASSERTED_PRINCIPAL=truein non-enforcing posture) and are explicitly marked unverified (verified=False,cage.principal.verified=false,x-cage-principal-verified: false) in permissive postures (dev,test,ci). On the GEAP Agent Gateway path,AgwAdapter.evaluate_request_authz()now fails closed (UNAUTHENTICATED, HTTP 403) whenrequire_verified_userholds andtri_identity.user_verifiedisFalse, andvalidate_tri_identity_binding()requiresuser_verified=TrueonTriIdentityCredentialBindingbefore releasing delegatedID-3OAuth credentials. - Evidence custodian recovers a lost custody cursor from WORM attestations (
fix(compliance), #519 — reported by Daniel Fairbanks (@danofairanks),src/compliance_bridge/evidence_custodian.py,tests/test_evidence_custodian.py,docs/architecture/EVIDENCE_CHAIN.md):EvidenceCustodianstores<stream_key>:custodyin the same Redis database as the evidence stream, so a full Redis loss (FLUSHDB, failover to an empty replica, volume loss) wiped both the stream and the cursor and caused_check_continuity()to treat a newly started genesis chain as a clean first-ever start. When the Redis cursor is absent,_recover_cursor_from_cold_store()now enumeratesevidence-stream/viaEvidenceColdStore.list_keys(), validates every attestation forstream_key(enforcingassert_citable()on.attestation.jsonand failing closed withEvidenceCustodyIntegrityErroron malformed JSON, schema, sequence ranges or stream IDs), and selects the latest custodied batch across chains using declaredCHAIN_ROTATIONlineage and(last_stream_id, last_sequence)ordering. On a full Redis wipe followed by a new chain, custody reads from the start of the new stream (min="-", including when a new chain starts in the same millisecond) and recordsGAP_KIND_CHAIN_ROTATION(or preservesprevious_chain_id,previous_last_sequenceandprevious_last_record_hashalongsideexpected_sequence: 0if the new chain's genesis was already trimmed); on cursor-only loss on the active chain, it restores the cursor in Redis and resumes afterlast_stream_idwithout duplicate batches or missed trim gaps. - Single-use Redis operations are retry-safe (
fix(gateway); found while fixing #511, reported by Daniel Fairbanks (@danofairanks)). Pooled clients frombuild_async_redis()retryConnectionErrorandTimeoutError, so when a reply is lost redis-py re-sends a command the server already ran, and a single-use write then saw its own first success as a conflict. Each write now stores a random per-attemptop_id(redis_single_use.py); a re-send that finds its ownop_idreports the original success, and every other writer, including a genuine replay, is refused exactly as before. Fixed: the routing-seal consume script andrevoke_seal()(a consumed seal read as a replay; a revocation that took effect returnedFalse);DeferQueue.park(), now one atomic script (a re-sentHSETNXleft a token with no status, TTL or expiry-index entry, so it could never be approved or expire); the revision CAS behindapprove()and_resolve()(a quorum approval returnedNOT_FOUND, an injection resolved the ticket but dropped itsinjection_data, an expiry skipped the DLQ);atomic_resolve()(a spent approval was refused as a replay);ConsequenceAuthorityStore.consume_once(), whose marker is now<binding>:<op_id>(get_binding()still returns the binding); and the DPoPjticlaim indpop_validator.py. The GCRA limiter inrate_limit.pystill charges a re-sent request twice: that only ever throttles more, and is documented. Tests:tests/test_redis_lost_reply.py. - Evidence stream appends are retry-safe; restore and custody fail closed on truncation (
fix(governance), #511 — reported by Daniel Fairbanks (@danofairanks)): The pooled async Redis client re-sends a command whose reply was lost, andEvidenceStreamSinktreated the re-sent compare-and-append'sCONFLICT(its own batch was now the head) as a lost race and re-sealed the same batch at the next sequence, writing duplicate records that still hash-verified._APPEND_BATCH_SCRIPTnow also receives the batch's lastrecord_hashandtimestamp_utcand answersALREADYwith the batch's entry IDs when that record is the head. On a realCONFLICTthe sink first reads thekentries after the record the batch was sealed against and resolves from them if the batch had landed before another writer appended. AConnectionError/TimeoutErrorre-sends the identical ARGV (_MAX_SETTLE_ATTEMPTS) instead of re-sealing, and a batch whose outcome stays unknown fails with the newEvidenceAppendUnsettledError(ingest_sync()refuses to seal; best-effortingest()returnsNone). The unused single-record_APPEND_SCRIPTis removed._restore_chain_state()now verifies the head's own hash withverify_record()and tracks the last acknowledged head: a sink that already holds a chain raisesEvidenceChainCorruptErroron an emptied stream, a head on another chain, or a head behind (or rewritten at) its last acknowledged sequence instead of silently starting a new chain, while a fresh sink on an empty stream still cold-starts.EvidenceCustodianrecords a new chain at genesis after another chain was custodied as aCHAIN_ROTATIONgap (CRITICAL log,cage_evidence_custody_gaps_total, previous chain id / last sequence / last record hash in the attestation) instead of a clean rotation, andCustodyVerifierreports it as a declared gap (DeclaredGap.kind,missing_to: None) that blocks citation unless explicitly allowed. OSCAL AU-9 / AU-10 statements updated insp800-53-component-definition.yaml. - The advisor's end-user principal is marked as asserted, not authenticated (
fix(advisor)!, #516 — reported by Daniel Fairbanks (@danofairanks),src/governed_financial_advisor/models/query.py,src/governed_financial_advisor/models/principal.py,src/governed_financial_advisor/infrastructure/gateway_client.py,src/governed_financial_advisor/server.py,src/gateway/governance/ingress/agw_adapter.py):CAGE_API_KEYauthenticates the calling client, not a user, so theuser_idthe advisor forwards asx-cage-principalis the caller's assertion.QueryRequest.user_idno longer defaults todefault_user: it is required and must be 1–64 characters matching^[A-Za-z0-9._:@/-]+$, the gateway's principal charset and cap.GatewayClient._build_headers()now passesx-geap-user-idthrough the sameprincipal_header_value()sanitizer and 64-character cap asx-cage-principal; it was forwarded raw. The/agent/queryspan recordscage.principal.verified=falsenext toenduser.id. On the Agent Gateway path,GeapTriIdentityContextgainsuser_verified, true only whenID-1comes from theact_sub/user_subclaim of a verified OIDC token, and records it asid1_user_verifiedin the envelope claims; no authorization decision readsID-1. The new Principal Trust Model section states the boundary; binding the principal to an authenticated identity is a follow-up. Breaking:POST /agent/queryreturns HTTP 422 whenuser_idis missing, empty, longer than 64 characters, or outside that charset. - Routing-verdict reuse binds the full order, and the advisor's trade previews cover every trade step (
fix(advisor), #513 — reported by Daniel Fairbanks (@danofairanks)):trade_contract.pymatched plan steps by substring, sobuyer_survey,executed_reportandbestseller_scanwere trades that failed with "Invalid trade amount None" and counted towardMAX_CONSECUTIVE_DENIALS. A step is now a trade only when its action is exactly one ofTRADE_STEP_ACTIONS(execute_trade,trade,buy,sell, ignoring case).extract_trade_step()returned only a plan's first trade step, sosafety_check_nodereported APPROVED and reset the denial budget on step 1 alone.extract_trade_steps()returns every trade step: the safety node submits each toPOST /governance/validate-action(a refusal of any step blocks the plan and counts one denial; a deferral parks it unless a later step is refused), andevaluator_nodedry-runs each. A missingconfidencewas previewed as1.0whileexecute_trade_actiondefaults it to0.0; it is now previewed as0.0, and a boolean, non-numeric or non-finiteconfidenceis rejected.canonical_trade_digest()bound only part of the order, so a cachedALLOWwas reused for a call with a differentlimit_price,account_id,order_type,venue,agent_idortransaction_id. The digest now binds the exactconfidenceand refuses any field outside the canonical payload (other than thetool_name/actionkeysgateway_tool_guardadds), so such a call is always re-validated, and only a single-trade plan answeredALLOWis cached. Tests:tests/test_trade_contract.py. - MCP tool calls are no longer resent after an unknown outcome (
fix(gateway)!, #514 — reported by Daniel Fairbanks (@danofairanks)).GatewayMCPClient.call_tool()caught any exception fromsession.call_tool, reconnected and sent the same arguments again, so a trade whose response was lost (read timeout, connection reset) ran a second, fully governed time. It now retries once only when nothing was sent (the session could not be established) or when the gateway's MCPToolAnnotationsmark the toolreadOnlyHintoridempotentHint. Otherwise it raisesMCPToolOutcomeUnknownError, chained from the transport error, and the caller must reconcile. ACallToolResultwithisErrorset raisesMCPToolExecutionErrorinstead of being returned as text, and is never retried. A session whose transport has closed is re-established before a request is sent. The gateway tools declare their hints (tool_provider.py,mcp_tool_server.py): market status and sentiment andsimulate_governance_checkare read-only,trigger_safety_interventionis idempotent, andexecute_trade_actionis destructive and not idempotent. The evaluator tools (agent.py) map both errors to theirERRORresults. The STPA saga template (stpa_compiler.py) now sends only the state keys a saga lists in the new, requiredlanggraph_saga.mcp_arguments(UCA-4:symbol,amount,currency,side,transaction_id; it used to send fields the tool does not take, such asapproval_tokenandaccount_id), parses the tool's text answer instead of calling.get()on it, records the request in thePENDINGintent, and leaves that entryPENDINGfor reconciliation on an unknown outcome (saga_nodes.pyregenerated). The advisor's unused trade wrappers (execute_tradeandpropose_trade) and their test are removed. Breaking:call_toolraisesMCPToolExecutionErrorfor anisErrorresult instead of returning its text, and an STPA saga withexecution_type: mcp_toolmust declaremcp_arguments. - A trade executes at most once per
transaction_id(fix(governance)!, #514 — reported by Daniel Fairbanks (@danofairanks)).transaction_idwas documented as an idempotency key, but nothing deduplicated on it: the clearance nonce was random and the Alpaca order carried noclient_order_id, so a resubmittedexecute_trade_actionwas governed and actuated again. The new kernelExecutionLedger(governor.components.execution_ledger, built byassemble_governor()over the gateway's shared Redis client) claimscage:exec:{action}:{caller_identity}:{key}withSET NX EXand a per-attempt token, and recordsEXECUTED,BLOCKEDorUNKNOWNwith a Lua compare-and-set on that token. Under an enforcing posture without Redis it refuses; DEV, TEST and CI fall back to a process-local registry.execute_trade_actionnow requires a UUIDv4 or UUIDv5transaction_id, claims it before governance, answers a resubmission with the original outcome without governing it again, or raisesUnsettledDuplicatewhile the first attempt is in flight or its outcome is unknown (either refusal is recorded in the evidence stream through the newrecord_refusal()ingovernance_middleware.py), records the outcome on every exit, and keeps the claim after anUNKNOWNactuation pending reconciliation. A resubmission that carries an approval (deferred_id) may re-claim a key whose earlier attempt wasBLOCKED, so a deferred trade still redeems its approval under the sametransaction_id; the gateway verifies and spends the approval before anything runs, and a key whose trade may have executed is never claimed again. Keys are namespaced by action only, because the tool is also reached through the gateway's HTTP/tools/executedispatch, which carries no request context.TradeOrderaccepts UUIDv4 and UUIDv5 transaction ids, the Alpaca order carriesclient_order_id(trade_executor.py), and the advisor'sgateway_tool_guardreplaces everyexecute_trade_actioncall'stransaction_idwith a UUIDv5 derived from the LangGraph thread and the tool call id, so a re-run of the same call resubmits the same key; the guard previews the order without that key, so a previewed routing verdict stays reusable.execute_trade_actionis now annotatedidempotentHint, soGatewayMCPClientmay resend it once after a send-phase failure, and the ledger answers the resend. Breaking:execute_trade_actionrefuses a call without a UUIDv4 or UUIDv5transaction_idinstead of generating one. - The trade policy's role is bound to the caller's verified workload identity (GHSA-w39v-5v7h-fxg9, CVSS 5.3 Medium) (
fix(security)!,src/cage_finance/tools/trader_binding.py,src/cage_finance/tools/trade_inputs.py,src/gateway/server/workload_identity.py).trader_roleandtrader_idwere arguments of theexecute_trade_actionMCP tool that the governed trader and the tool guard forwarded unchanged totrade_governance.rego, which sets a trade's limits by role, so a prompt-injectedseniorlifted the unreviewed limit from $5,000 to $500,000 on the committing trade path.TradeInputResolvernow owns both keys and binds them on every path that evaluates a trade (validate-action previews, the committing run, POST_HITL re-runs) from the identityWorkloadIdentityMiddlewareverified for the request, read through the new kernel accessorcurrent_client_identity().CAGE_TRADER_ROLE_BINDINGS(a JSON object) maps a verified identity tojuniororsenior, optionally with atrader_id; an unbound identity isjunior,trader_iddefaults to the identity, and an invalid configuration stops the finance plugin from assembling. Without a verified identity an enforcing posture binds no role, so OPA denies; dev, test and CI bindjunior. Breaking:execute_trade_actionand its MCP schema no longer taketrader_roleortrader_id(a/tools/executecall that still passes them fails with an error before any governance runs, so nothing executes), and the advisor no longer sends a role:trade_params(),TRADE_PARAM_KEYSand the advisor'sExecutionPlandrop it (socanonical_trade_digest()binds no role and refuses a payload that still namestrader_roleortrader_id), andDEFAULT_TRADER_ROLEis removed. STPA control-action params gain a genericserver_bound: trueflag, and a saga whosemcp_argumentsnames a server-bound param of its forward action fails control-structure validation, so no generated saga tool call can send one. The advisor'sgateway_tool_guardnow strips a model-supplieddeferred_idfrom every executed tool call, not only from the validate-action params; only the approved resume adds one. Found during maintainer triage. - The OSCAL exporter keeps hand-maintained FTRA statements (
fix(compliance)—src/gateway/governance/oscal_ssp_exporter.py,tests/test_oscal_ssp_exporter.py). Eachexportreplaced the FTRA component ofcomponent-definition.yamlwholesale, so the committed SI-10 statement, which names the domain's UCA rules and OPA policy files, was overwritten by the kernel's seed text, which may name no domain path (Gate G3)._apply_ftra_component_patch()now keeps each existing statement's text, matched by implemented-requirement UUID, and replaces only the component's structure and machine fields such aslast-compiled; a statement with no text is seeded, so deleting adescriptionre-seeds it. The seed sentence no longer reads "STPA UCA rules (domain STPA UCA rules)".
Changed
- Skill certification receipts bind the evaluation stages and evidence (
cage.skill.cer.v2) (feat(compliance)!, #512 — reported by Daniel Fairbanks (@danofairanks)). A stage-0-only.cerreceipt signed a mean composite lift of0.0, a hard-codedmodel_idandevals_sha256 = sha256(b""), and the defaultGcsSkillCertificationVerifieraccepted it without theSKILL.mdbytes.SkillCertificationReceiptnow signsstages_run, anevidence_level(CerEvidenceLevel:STATIC_ONLY,PAIRED_TRACESorLIVE_ROLLOUT),baseline_traces_sha256andwith_skill_traces_sha256oradapter_refandadapter_config_sha256, andnulllift,model_idandevals_sha256values when no lift was measured. Construction andfrom_dict()fail closed on any field combination that contradicts the declared stages or evidence level.GcpSkillCertifier.create_and_sign_receipt()requiresstages_runandevidence_level, computes every evidence digest itself, and takesmodel_idfrom the lift report (the rollout bundle's model for adapter runs) with no default.cage-skill-eval(cli.py) signs the stages and evidence it actually evaluated and adds--model-id, required with--baseline-traces/--skill-traces. The gate, the verifier andGovernedGEPAOptimizersharelift_meets_threshold(): a0.0minimum requires strictly positive lift, andNaNor unmeasured lift never passes. The verifier refuses every other schema version (unsupported_schema_version) andSTATIC_ONLYreceipts unless constructed withallow_static_only=True(static_only_receipt).verify_and_transition_gateway_mode()is nowasync, requires an asyncCerArtifactVerifierand at least one artifact, and no longer falls back to astatus == "CERTIFIED"presence check. Breaking:cage.skill.cer.v1receipts are no longer accepted (re-certify to emit v2 receipts);verify_cer_receipt()requires the keyword-onlyexpected_skill_md_bytes;create_and_sign_receipt()requiresstages_runandevidence_leveland no longer defaultsmodel_id;run_certification_gate()requiresmodel_idwith trace files; andverify_and_transition_gateway_mode()is a coroutine without thecer_verifiedflag.
Other changes
Merged in this release without an entry of their own:
- Correct provider_06 verify-release wire headers (
fix(governance), #469) - Reject actuator_01 receipts not bound to the envelope (
fix(governance), #479) - Keep cumulative-spend keys in sync in eval trajectories (
fix(governance), #483) - Fail closed on NARROW and unknown verdicts in
CageClient(fix(gateway), #484) - Remove vendor aliases and add IPR notice (
refactor(governance), #508) - Harden kernel core validation and governance controls (
fix(governance), #533) - Align partner v0.3 contracts and harden infra isolation (
feat(governance), #534) - Resolve live GKE e2e telemetry and defer bugs (
fix(gateway), #535) - Restore the quickstart compose bootstrap and add a smoke gate (
fix(infra), #537) - Dependencies: upgrade locked dependencies and base images for the CI scans (#509); bump certifi (#510); pin filelock 3.20.3 and raise setuptools to 84.1, which clears the pip-audit and Trivy image-scan findings (#543)
- Documentation: #518, #532, #536, #538, #539, #540, #542; CI: #485
Full changelog: v3.2.0...v3.3.0