status.md

Project Status

Baseline record — not a specification. This document describes as-built or operational state; the normative target is the specification set indexed in docs/spec-policy.md, and where this document conflicts with a specification home, the specification wins.

6529Stream is pre-audit and not production-ready.

The current Gate A smoke baseline proves:

• The checked fresh-contributor path in docs/first-30-minutes.md covers pinned tool expectations, forge not being on PATH, Windows PowerShell wrapper usage, known warning noise, generated artifact drift, docs-only paths, Solidity/test paths, and no-secret maturity boundaries.
• The checked GitHub issue-template surface captures integration reports, public-safe audit finding intake, and release evidence requests while preserving pre-audit, not-production-ready, no-secret, and private security reporting boundaries.
• The checked PR-template surface requires roadmap/gate linkage, validation evidence, release-impact classification, generated-artifact impact, and breaking-change approval references before review.
• Foundry is configured to compile smart-contracts.
• forge build runs against Solidity 0.8.19.
• forge build --sizes --via-ir --skip test --skip script --force remains an aggregate diagnostic and warning-collection build. Canonical release bytecode instead comes from python scripts/build_release_artifacts.py, which compiles each configured source and import closure in isolation, rejects test/script source units in configured targets, retained build-info compiler inputs, and artifact metadata, validates a deterministic input/artifact receipt, projects only Forge's exact absolute worktree path carriers to a checked stable . / lib representation so receipt hashes are portable across otherwise identical checkouts, and uses staged replacement with rollback on caught failures to restore configured artifacts plus retained compiler inputs and the receipt to dedicated ignored out-release/. Ordinary Forge builds and scripts continue to use out/. python scripts/check_contract_size_budget.py checks those canonical artifacts against EIP-170 and enforces the interim StreamCore development floor from release-artifacts/contracts.json. This closes the local release and verification evidence gap tracked by issue #674 and makes the restricted-source boundary explicit under issue #675 without making Forge deployment scripts or broadcasts consume the canonical initcode. That deployment binding remains a production blocker under issue #677, so the pre-audit, not-production-ready posture is unchanged. The committed release-artifacts/latest/bytecode-release-proof.json records the permanent, target-isolated StreamCore production runtime as 18,997 bytes, leaving 5,579 bytes of EIP-170 headroom under the IR-optimized deployment profile. This is 3,579 bytes inside the non-waivable 2,000-byte deployment margin, 3,187 bytes below the approved 22,184-byte objective, and 5,131 bytes smaller than the 24,128-byte transitional operation-identity build. Historical 21,792-byte scratch and merged measurements remain non-additive; only this complete linked build is the current size proof. The permanent Core now includes the locked mint, metadata, entropy, governed-parameter, pointer, and restricted ERC-4906 seams rather than relying on a temporary size exception. Accepted ADR 0018 is now implemented in the current pre-genesis source: one manager batch root is replay-protected in the ledger, exactly one unique operation ID identifies each token transition, derivation and nonce reservation precede ledger consumption, and Core retains only its current prepared-pair equality lock. Exact typed primary settlement and repeated-sale replay remain ADR 0019 / #694 blockers; operation-root uniqueness does not close them. The target restricted ERC-4906 single/batch refresh emitters required by #667 are implemented and included in the measured Core. The final recovery boundary is IStreamFinalityRecoveryCore: lastAllocatedTokenId() (0x254b22bc) XOR emitBatchMetadataUpdate(uint256,uint256,bytes32) (0x908c18bd) produces ERC-165 ID 0xb5c73a01. Core advertises IERC165 and that exact ID, rejects 0xffffffff, and implements the real batch emitter; fallback-only handling remains nonconformant. Issue #667 owns the separate fail-closed interface probe and registry consumer. The stricter bytecode-spend ceiling remains the reviewed 22,184-byte approved baseline with 2,392 bytes of baseline margin. The architecture policy in docs/architecture.md#product-extension-and-size-budget-policy makes future collector/product surfaces satellite-first by default through satellite contracts, read adapters, linked libraries, release artifacts, or documentation-only evidence unless Core ownership/security invariants require otherwise.
• The strict release-mode profile now compares the current implementation catalog with the canonical exhaustive genesis profile and fails closed on missing, extra, ambiguous, wrong-scope, interface, marker, and fallback entries. ADR 0017 reduces the canonical launch profile to 37 entries and forbids probe rows or bindings. The catalog remains incomplete, and its singleton-oriented manifest model cannot yet prove every required distinct deployment instance. This keeps production blocked under issue #656.
• The exact ADR 0017 identifier surface is checked at 22 GGP plus three GTP rows, and the schema-validated release-artifacts/governed-parameter-inventory.json now pins the exact launch policy, 50-binding host-profile coverage, failure classes, sizing/cadence evidence requirements, and fixed-stipend review obligations. The committed inventory is intentionally incomplete: only generic store hosts currently implement registration, and concrete candidate host addresses, genesis values, immutable floors, reviewed evidence digests, and fixed-stipend compatibility remain unavailable. Current guarded-consumer lists are also explicitly planning rather than exhaustive. The ordinary checker accepts that honest planning state; production release invokes --require-complete and rejects it. It also rejects every self-reported complete candidate until #656 supplies structured instance and linked-library reconciliation, and rejects self-reported complete evidence until #684 adds candidate-instance-bound measurement/cadence, reproduction, and reachable raise-chain semantics. This independent production blocker is retained as RISK-GOV-004 and tracked by issue #684. Issue #671 now binds the reusable admission source to the permanent Core's three actual read paths, hostile threshold/returndata/full-budget tests, six via-IR completion measurements, actual Core below/at/above regressions, and the planning floor/genesis tuple in the governed inventory. Issues #656/#670/#684 still block exact candidate, fixed-stipend, cadence, rehearsal, and independent-review evidence.
• python scripts/test_solidity_formatting.py and python scripts/check_solidity_formatting.py enforce the scoped Solidity formatting policy: 34 formatting-required first-party/provider files pass forge fmt --check, and the raw all-files diagnostic is allowed to fail only for the 17 documented vendored/provenance formatting exemptions.
• python scripts/test_warning_dispositions.py, python scripts/run_forge_size_log.py --log cache/forge-size.log, and python scripts/check_warning_dispositions.py --solc-warnings-log cache/forge-size.log enforce the checked warning disposition baseline in docs/warning-dispositions.md. That baseline records fixed NatSpec warning cleanup and accepted solc, documentation, linter, vendored, test-only, ABI-compatibility, and StreamCore size-tradeoff warning rows without treating warning quietness as protocol correctness proof.
• python scripts/test_natspec_coverage.py and python scripts/check_natspec_coverage.py enforce the checked NatSpec coverage baseline in docs/natspec-coverage.md and release-artifacts/baselines/v0.1.0/natspec-coverage.json. The current baseline is a burn-down queue for missing release-surface NatSpec, not proof that API documentation is complete.
• forge test -vvv executes real tests for admin guards, target-scoped function-admin permission regressions, domain-scoped pause controls, signer-manager lifecycle controls with approved targets, EIP-712 and ERC-1271 drop authorization, auction custody and payment credits, fixed-price pull-payment credits, curator reward claim credits, and randomness lifecycle behavior. Current emergency-withdrawal target-state tests also cover explicit emergency recipients, StreamMinter surplus withdrawal, NextGenRandomizerRNG reserve boundaries, request-level randomizer reserve lifecycle accounting, dependency-script segment-safe content hashing, explicit local-initialization regressions, vendored OpenZeppelin utility-library provenance/behavior regressions, and retained airdrop mint-accounting behavior after removal of dead public/allowlist counters.
• Focused randomizer reserve lifecycle tests now cover arRNG request-cost spending, multiple pending requests, fulfilled requests, stale requests, failed post-processing, deterministic retry, forced ETH, and unauthorized emergency-withdrawal attempts while proving remaining provider reserves stay represented by totalRandomnessReserved(), totalOwed(), and totalReserved() with zero emergency-withdrawable surplus.
• Randomizer operations evidence now has a schema, local Anvil evidence bundle, and local/CI checker. The local bundle records deployed adapter addresses, placeholder provider addresses, provider epochs, local funding status, lifecycle-control evidence, reserve-accounting evidence, retained artifacts, and redaction policy. Fork/testnet/live provider funding and request-health evidence remains future Gate E release work.
• Payment sequence fuzzing now covers mixed fixed-price mint, auction bid, auction settlement, curator claim, withdrawal, emergency withdrawal, randomizer reserve, and forced-balance operations, proving the current local ledgers keep category totals, owed totals, balance coverage, reserves, and emergency-withdrawable surplus coherent after each step. Current value-holding payment surfaces also expose ADR-style local-ledger aliases for reserved balances and surplus where applicable.
• Supply/replay/freeze sequence fuzzing now covers mixed fixed-price mints, drop cancellations, consumed-drop replay attempts, cancelled-drop mint attempts, burns, metadata edits, freeze attempts, stored freeze-manifest stability, final-supply tightening, and post-freeze mint/burn/token-data rejection. The invariant reasserts global and collection live supply, minted-ever counters, burn counters, burn audit state, and consumed/cancelled drop ID state after each step.
• Auction consistency sequence fuzzing now covers mixed signed auction-drop registration, escrow custody, cancellation, first bids, higher outbids, underbid and late-bid rejection, no-bid and with-bid settlement, repeat settlement attempts, bidder/proceeds withdrawals, emergency surplus, and forced-balance operations. The invariant reasserts known token custody, active highest-bid escrow, previous-bidder credits, settlement proceeds credits, terminal ownership, failed invalid-operation preservation, and auction-local totalOwed()/totalReserved()/surplus() view coherence after each step.
• Randomizer tests now cover request lifecycle views, callback validation, raw-output hash storage, failed post-processing state, bounded deterministic post-processing retry, and the conservative provider-migration policy that blocks lifecycle-aware provider replacement while collection requests are pending. They also prove RandomizerNXT cannot be configured as a production randomizer after removal of the concrete weak XRandoms helper from production source.
• Randomizer/admin stateful coverage now composes arRNG reserve funding, request costs, unique token requests, fulfillment, stale marking, failed post-processing, retry success/failure, provider and epoch replacement attempts, randomness-request pauses, token-collection drift, and emergency-withdrawal calls in bounded generated sequences while reasserting lifecycle indexes, pending counters, terminal request fields, retry limits, core token hashes, and zero-surplus reserve views. This is local adversarial coverage only; fork/testnet/live provider operations evidence remains future release work.
• Auction/drop/arRNG composition tests now execute signed auction drops through the real drop, auction, core, and arRNG contracts, proving randomness-request pauses roll back drop execution without consuming the authorization, pending arRNG requests block randomizer migration without auction drift, and auction settlement before fulfillment preserves winner custody, credits, total owed values, and token/request binding after later fulfillment. This remains local adversarial coverage only; it does not replace fork/testnet/live provider operations evidence.
• The same auction/drop/arRNG composition suite now covers post-execution signer lifecycle controls: signer-epoch invalidation, signer rotation, replay attempts, consumed-drop cancellation attempts, and drop-execution pauses after auction creation leave existing auction custody, bid accounting, pending request bindings, settlement, and later arRNG fulfillment coherent. This remains local adversarial coverage and does not replace production signer, Safe, or provider operations evidence.
• The auction/drop/arRNG composition suite now covers no-bid and cancellation-like terminal paths: no-bid settlement before fulfillment, contract-poster pending no-bid claims, pre-bid auction cancellation, cancelled signed auction authorizations, and rejected terminal repeats preserve custody or claimant state, zero-accounting boundaries, pending request bindings, and later arRNG fulfillment. This remains local adversarial coverage and does not replace fork/testnet/live provider operations evidence.
• Fixed-price/drop/arRNG composition tests now execute paid fixed-price drops through the real drop, minter, core, and arRNG contracts, proving randomness-request pauses roll back drop execution without consuming the authorization or crediting payments, post-execution signer lifecycle controls do not disturb fixed-price credits or pending request bindings, and poster/protocol withdrawals before arRNG fulfillment do not break later fulfillment. This remains local adversarial coverage and does not replace production signer, Safe, or provider operations evidence.
• Fixed-price and auction arRNG request-ID collision tests now force a reused provider request ID through the public signed-drop paths, proving the second drop reverts with RandomnessRequestAlreadyExists without consuming the authorization, minting or registering a token, changing supply or drop counts, creating payment credits or auction state, changing pending request accounting, or breaking fulfillment of the first token's request. This remains local adversarial coverage and does not replace fork/testnet/live provider operations evidence.
• Burned pending arRNG composition tests now cover paid fixed-price and settled auction drops that burn before fulfillment, proving payment/proceeds accounting, consumed-drop state, request bindings, burned-token audit randomness, no-metadata-update fulfillment, freeze eligibility, and frozen manifest stability remain coherent. This remains local adversarial coverage and does not replace fork/testnet/live provider operations evidence.
• Metadata tests now prove dependency chunk boundaries are included in typed content hashes while preserving the existing rendered generative script output. StreamDependencyRegistry.t.sol proves immutable dependency version records, provenance/deprecation views, collection dependency key/version/content-hash/registry-address pins, explicit repinning, and output stability after later registry versions or registry swaps. StreamMetadataGolden.t.sol locks current off-chain pending/stale/failed/final URIs plus schema-v1 on-chain pending/stale/failed/final base64 JSON data URIs against fixtures. The on-chain schema exposes metadata_schema_version and metadata_state; pending, stale, and failed metadata omit final animation HTML while the token hash remains zero; lifecycle lookup failures fall back to pending; and final token hashes override lifecycle display state. StreamMetadataEvents.t.sol proves ERC-4906 interface support and current metadata update event semantics for token-level updates, and it now proves optional ERC-721 Enumerable support is not advertised while the contract preserves a live totalSupply() view. collection-range updates, randomness fulfillment, mint-only paths, and burn. StreamMetadataFreeze.t.sol proves the current collection freeze boundary: ended mint window, elapsed final-supply delay, ADR 0006 rejection of live tokens whose randomness remains pending, stale, or failed instead of final, stored manifest hash/event, final supply tightening, and post-freeze rejection for current StreamCore metadata-significant mutation paths. StreamCoreBurn.t.sol proves burn metadata semantics: burned tokens lose ownership and tokenURI availability, emit protocol burn audit events, retain audit state, exclude burned supply from live collection supply, reject remint attempts for previously burned token IDs, and allow valid VRF/arRNG post-burn fulfillment to record audit-only randomness without ERC-4906 metadata updates or freeze-manifest changes. StreamMetadataEscaping.t.sol proves the current on-chain JSON string escaping and raw attribute schema guard: escaped collection/image fields decode to parseable JSON, brackets inside quoted attribute values remain valid, valid JSON string escapes are accepted, and breakout/control-character/unterminated/malformed semantic attribute fragments revert. It also decodes final animation HTML and proves the wrapper escapes the external library attribute, embeds tokenData and dependency script content through escaped JavaScript strings, and neutralizes closing-script sequences so hostile metadata inputs do not create extra raw wrapper </script> tags. StreamMetadataSizeLimits.t.sol proves byte caps for collection display fields, collection script chunks and counts, token data, token images, token raw attributes, generated tokenURI output, dependency script chunks and counts, and dependency provenance strings. StreamMetadataUriPolicy.t.sol proves the renderer content/script URI policy helpers and the production URI guards for allowed https://, ipfs://, and ar:// content URIs plus rejected empty required token image URIs, JavaScript, hostless HTTPS, whitespace-bearing token image/base URI inputs, and non-HTTPS external library URLs. StreamMetadataUtf8.t.sol proves the shared strict UTF-8 scanner accepts valid ASCII and multibyte sequences, rejects invalid lead/continuation, overlong, surrogate, out-of-range, and truncated sequences; enforces that DependencyRegistry rejects invalid UTF-8 dependency scripts/provenance; and enforces that StreamCore rejects invalid UTF-8 collection fields, collection script chunks, token data, token image URIs, and token raw attributes while preserving size-before-UTF-8 error ordering. StreamRandomizerLifecycle.t.sol also proves unsupported lifecycle providers do not block randomizer migration, while lifecycle-aware providers whose pending-request probe fails still block replacement. StreamCoreCustomErrors.t.sol proves the typed failure selectors used to recover StreamCore bytecode headroom for function-admin authorization, artist-signature authorization, metadata-array length validation, and final-supply timing, plus missing collection data rejection before final supply math and the compact TokenNotMinted() selector used by Core metadata views. scripts/test_metadata_fixtures.py and scripts/check_metadata_fixtures.py validate the committed metadata golden fixtures outside Foundry by strictly decoding JSON and HTML data URIs, parsing metadata JSON, rejecting invalid UTF-8 fixture payloads, validating semantic attribute shape, checking current URI scheme policy, and asserting the generated final animation wrapper has exactly the expected script boundaries. scripts/test_metadata_browser_sandbox.py and scripts/check_metadata_browser_sandbox.py execute the committed final on-chain animation fixture in Chromium inside a sandboxed iframe, serve the expected external dependency through a deterministic stub, fail unexpected outbound HTTP(S) requests, assert the bootstrap values, capture page/console errors, and prove parent-document access is blocked. scripts/test_rehearsal_metadata_browser_sandbox.py and scripts/check_rehearsal_metadata_browser_sandbox.py execute metadata generated by a local deployment rehearsal: the Forge script deploys the local stack, mints through the EIP-712 drop authorization path, finalizes randomness/image/attribute inputs, returns the generated on-chain tokenURI, and the Python checker runs that generated final animation through the same Chromium sandbox policy.
• StreamDeploymentManifest.t.sol proves the first Gate E local deployment rehearsal can deploy and wire the stack, configure sample admin/pause/signer ceremony state, transfer Ownable control to the configured Safe placeholder, revoke the temporary deployment admin, parse the deployment manifest schema/example JSON artifacts, and run a local auction ceremony from signed auction drop through active auction custody, bid escrow, with-bid settlement, poster/protocol/curator proceeds withdrawals, and zero owed funds. It also proves a local emergency redeployment rehearsal with distinct old/replacement deployment versions, manifests, drop EIP-712 domains, core, drops, and auction addresses, Safe-rooted admin ceremony state on both deployments, temporary deployer-admin removal, and a replacement fixed-price mint smoke path. The suite-level test proves script/RehearseDeploymentSuite.s.sol returns deployment, auction, and emergency result groups plus a combined suite hash.
• forge script script/RehearseDeploymentSuite.s.sol:RehearseDeploymentSuite --sig "run()" --via-ir executes the three local rehearsal flows as an aggregate suite-level evidence command. The local/CI smoke gate also runs the three standalone forge script entrypoints, keeping the original script execution contexts automated while retaining one combined suite result for release-gate review.
• scripts/test_deployment_rehearsal_gate.py and scripts/check_deployment_rehearsal_gate.py statically guard the aggregate suite command, all three standalone rehearsal commands, and the CI retained log names across Make, Bash, PowerShell, and CI before the Forge rehearsal scripts execute. This is wiring-parity evidence for the local gate, not fork, testnet, or live deployment evidence.
• scripts/test_release_artifacts.py and scripts/generate_release_artifacts.py --check prove the committed release-artifacts/latest/ baseline matches current Foundry ABI/event output, including ABI checksums, bytecode checksums, standard/custom interface IDs, and event topic catalog entries.
• scripts/test_protocol_surface_report.py and scripts/generate_protocol_surface_report.py --check prove the committed release-artifacts/latest/protocol-surface-report.json matches the current production Foundry artifacts for functions, selectors, events, topic0 values, custom errors, ABI hashes, bytecode hashes, and runtime sizes.
• scripts/test_source_verification_inputs.py and scripts/generate_source_verification_inputs.py --check prove the committed release-artifacts/latest/source-verification-inputs.json retains production contract source hashes, compiler settings, constructor ABI, bytecode/linking status, and verification command templates from the current Foundry artifacts.
• scripts/test_dependency_artifact_manifest.py and scripts/generate_dependency_artifact_manifest.py --check prove committed dependency artifact descriptors under release-artifacts/dependencies/ resolve only to packaged dependency files, reject malformed keys and duplicate identities, and keep release-artifacts/latest/dependency-artifact-manifest.json current.
• docs/dependency-operations.md documents the production dependency ceremony: proposal, review, source packaging, registry registration, unfrozen collection repinning, deprecation, rollback by corrective version, frozen collection immutability, and source-retention evidence.
• scripts/test_abi_compatibility.py and scripts/check_abi_compatibility.py --check prove the current production contract and published interface ABI surfaces remain compatible with the committed release-artifacts/baselines/v0.1.0/abi-surface.json baseline. The first baseline fails on removed or changed functions, events, custom errors, constructors, fallback, or receive entries and reports additive entries as compatible.
• scripts/test_broadcast_manifest_input.py and scripts/generate_broadcast_manifest_input.py --check prove the sanitized Foundry broadcast fixture maps exactly to the expected Anvil deployment contract set, chain ID, transaction hashes, successful receipts, and non-duplicate deployed addresses without committing secret-like keys.
• scripts/test_deployment_manifest.py and scripts/generate_deployment_manifest.py --check prove the local Anvil placeholder manifest is generated from committed inputs, references current ABI/runtime bytecode hashes, and carries a deterministic manifest checksum. The same generator also checks the sanitized broadcast-derived manifest input with --config deployments/config/anvil-6529stream-v0.1.0-001-broadcast.json.
• scripts/test_address_books.py and scripts/generate_address_books.py --check prove the generated local Anvil address books are compact, deterministic projections of the committed placeholder and broadcast-derived deployment manifests plus release artifact contract metadata.
• scripts/test_ceremony_evidence.py and scripts/check_ceremony_evidence.py prove the committed no-secret local Anvil ceremony evidence bundle follows deployments/schema/ceremony-evidence.schema.json, references existing deployment manifests, address books, checksum inputs, rehearsal scripts, and verification status by SHA-256, rejects stale file hashes and secret-like keys, and reserves fork/testnet/live retained evidence contents for later Gate E ceremonies.
• scripts/test_release_manifest.py and scripts/generate_release_manifest.py --check prove the generated top-level release manifest ties the release artifact catalog, dependency artifact manifest, ABI compatibility baseline, deployment manifests, address books, ceremony evidence, public-beta evidence status, schemas, changelog, governance docs, and unavailable release-ceremony outputs together without drift.
• scripts/test_release_checksums.py and scripts/generate_release_checksums.py --check prove the committed release-artifacts/latest/SHA256SUMS and release-artifacts/latest/release-checksums.json bundle covers the current release artifact, dependency artifact source, broadcast fixture, deployment manifest, address-book, config, schema, public-beta evidence status, and release-manifest files.
• scripts/test_markdown_links.py and scripts/check_markdown_links.py prove the contributor, docs, ops, GitHub template, and release-artifact Markdown surfaces do not contain stale local file links, missing heading anchors, duplicate-anchor drift, repository escapes, or invalid line anchors.
• scripts/test_release_signatures.py and scripts/check_release_signatures.py prove the committed no-secret local release signature evidence follows release-artifacts/schema/release-signature-evidence.schema.json, records the self-referential release manifest/checksum bundle boundary, rejects stale retained hashes, symlinked evidence refs, and secret-like values, and reserves production detached checksum signatures, signed Git tags, signer identity, and verification output for public release ceremonies.
• scripts/test_non_local_release_evidence.py, scripts/check_non_local_release_evidence.py, scripts/test_marketplace_indexer_evidence.py, and scripts/check_marketplace_indexer_evidence.py fail future reviewed release evidence closed unless generic non-local retained artifacts, marketplace/indexer evidence envelopes, and retained Markdown references are ordinary repo-relative files, not symlinked files or files reached through symlinked directories.
• scripts/test_drop_authorization_signing_evidence.py and scripts/check_drop_authorization_signing_evidence.py prove the committed no-secret drop authorization signing evidence template follows release-artifacts/schema/drop-authorization-signing-evidence.schema.json, ties retained signing ceremony metadata to a generated unsigned EIP-712 payload, validates domain/message/digest fields, signer epoch, reviewer and signature metadata, retained artifact hashes, and path boundaries, and keeps production signing evidence blocked until reviewed non-local evidence exists.
• scripts/test_signer_custody_readiness.py and scripts/check_signer_custody_readiness.py prove the committed no-secret signer custody readiness template follows release-artifacts/schema/signer-custody-readiness.schema.json, validates custody owner, signer manager, signer epoch source, signer-service class, ERC-1271 support status, rotation/revocation readiness, monitoring/runbook references, reviewer metadata, retained artifact hashes, path boundaries, and no-secret policy, and keeps public-beta readiness blocked until reviewed non-local custody evidence exists.
• scripts/test_live_deployment_manifest_evidence.py and scripts/check_live_deployment_manifest_evidence.py prove the committed template for future live_deployment_manifest evidence is public-safe and fail future pending/reviewed evidence closed unless retained manifest files are live mainnet chain ID 1, have finalized nonzero contract addresses, retain bytecode hashes and constructor arguments, agree with the retained address book, avoid symlinked retained files, avoid symlinked retained-file directories and secret-shaped content, and match optional declared sha256: digests. Issue #227 remains open until real reviewed live manifest evidence is retained.
• scripts/test_live_metadata_browser_evidence.py, scripts/check_live_metadata_browser_evidence.py, scripts/test_production_broadcast_retention.py, and scripts/check_production_broadcast_retention.py share the retained-path resolver with the other release evidence checkers so future pending/reviewed live metadata-browser and production broadcast retention evidence cannot reference symlinked leaf files or files reached through symlinked directories. Issues #473 and #226 remain open until real reviewed evidence is retained.
• scripts/test_testnet_deployment_rehearsal_evidence.py and scripts/check_testnet_deployment_rehearsal_evidence.py fail future pending/reviewed Sepolia deployment rehearsal evidence closed unless retained transcript, broadcast, manifest, address-book, and gas/invariant files are ordinary repo-relative files, not symlinked files, and remain no-secret. Issue #217 remains open until real reviewed testnet deployment evidence is retained.
• scripts/test_public_beta_evidence.py and scripts/check_public_beta_evidence.py prove the committed no-secret public-beta evidence status follows release-artifacts/schema/public-beta-evidence.schema.json, keeps public beta and production release blocked while non-local audit, deployment, ceremony, randomizer, signature, signed tag, address, broadcast, and explorer evidence is missing, rejects stale retained hashes, rejects secret-like keys or values, and rejects ready claims while blockers remain.
• scripts/test_architecture_threat_model.py and scripts/check_architecture_threat_model.py prove docs/architecture.md and docs/threat-model.md retain the current auditor-facing component map, role boundaries, value/custody flows, threat categories, residual risks, evidence links, and pre-audit/no-production-claim maturity language before the audit package and release manifest are checked.
• scripts/test_audit_package.py and scripts/check_audit_package.py prove docs/audit-package.md remains a complete auditor-facing index over current maturity, scope, ADRs, invariants, the open Slither blocker, local deployment and release evidence, known blockers, accepted non-Slither local-baseline dispositions, and security reporting. The generated release manifest records the architecture, threat model, and audit package hashes as governance documents before the checksum bundle is refreshed.
• scripts/check_slither_baseline.py keeps a canonical normalized first-party production set in ops/SLITHER_BASELINE.json, checks its Markdown mirror and provenance without invoking Slither during the fast default gate, and runs a dedicated pinned exact-drift analysis in CI. The current set is 32 Open rows (2 High, 30 Medium): zero confirmed gaps, six design-review rows, and 26 pending dispositions. This is a release blocker and not a risk acceptance. Bounded assembly made the Governance Executor's proposal-selected native-value authority invisible to Slither without removing it. That analyzer disappearance is not remediation. The checksum-covered closed-world Governance V2 catalog now makes unknown action-class/target/selector tuples and unreviewed native value fail closed at scheduling and execution. RISK-GOV-003 remains a separate High open blocker under issue #658 until issue #656 supplies exact candidate bindings and the required deployment, rehearsal, monitoring, and independent-review evidence. The Governance V2 foundation remains pre-audit and not production-ready.
• scripts/test_release_readiness.py and scripts/check_release_readiness.py prove docs/release-readiness.md remains a Gate G dashboard that separates passing local evidence from missing fork/testnet/live evidence, production signatures, signed Git tags, verified deployed addresses, explorer verification, external audit, and post-audit remediation blockers before the release manifest and checksum bundle are refreshed.
• forge snapshot --match-path test/StreamGasSnapshot.t.sol --check release-artifacts/baselines/v0.1.0/gas-snapshot.snap proves the committed local gas snapshot still matches the focused Gate D operations for fixed-price mint, auction bid, auction settlement, curator claim, final on-chain tokenURI, and dependency/script reads.
• scripts/test_changelog_check.py and scripts/check_changelog.py prove release-impacting branch changes include a non-placeholder Unreleased changelog entry before they can pass the local/CI gate.
• CI can run the same build/test smoke commands and publish logs.

The current tests are regression tripwires, not a correctness proof. Known blockers remain tracked in ops/ROADMAP.md, including any future unified pull-payment ledger abstraction or protocol-wide aggregation layer, canonical randomizer lifecycle ownership, remediation or reviewed disposition of every open high/medium first-party production Slither row, fork/testnet deployment rehearsals, production manifest generation from live broadcast outputs, production detached checksum signatures, signed release tags, production address books, verified live deployment hashes and explorer submissions, remaining generated HTML/JavaScript render-sandbox hardening beyond the committed browser fixture and local deployment-rehearsal check, fork/testnet/live production metadata browser evidence, fork/testnet/live ceremony evidence contents, fork/testnet/live emergency redeployment evidence contents, fork/testnet/live randomizer operations evidence contents, deployment discipline, and fork/testnet/live invariant coverage. Fixture-level invalid UTF-8 regressions, dependency registry production UTF-8 guards, and Core-level production UTF-8 guards are now covered.

Contributor, security intake, architecture, threat-model, audit package, public-beta evidence, and release-readiness files exist so future work can be packaged and reviewed consistently, but they do not change the pre-audit status.