VSTD v1.2.0

Source tag signature status: SIGNED_AND_GITHUB_VERIFIED (valid).

Every uploaded asset is bound to the exact public commit by the external release manifest and a GitHub artifact attestation. The wheel and source distribution published to PyPI are the same tested files attached here. Verify a GitHub asset with:

gh attestation verify PATH_TO_ASSET --repo TimeLordRaps/verifier

Public surface and integrations

• Restructure the public first-view path around one bounded project description, one
  deterministic demonstration, one canonical maturity table, skeptical claim limits,
  contributor routes, and release/citation boundaries; align Pages and package metadata
  without changing normative or serialized-receipt semantics.
• Normalize the public architecture as a verification complex of named closure
  coordinates and cumulative numbered profiles; reserve VSTD-4 rung, candidate depth,
  verification order, compatibility level, and checker-cost tier for their distinct uses.
• Add experimental workflow profile 0.1 with deterministic canonicalization, strict
  validation, bounded work-allocation records, additive amendments and challenges,
  explicit unresolved horizons, and verdict-neutral platform events.
• Add a normalized GitHub adapter for issues, commits, workflow runs, artifacts, and
  pull requests. Successful workflows and merges retain verification_effect = NONE
  unless a separate native result is explicitly mapped through a bound VSTD receipt.
• Add vstd experiment validate and vstd experiment github-events as offline,
  verdict-neutral entry points. Repository artifacts are explicitly NOT_CHECKED with
  exit code 2 unless their root is supplied.
• Add a machine-readable schema, checked-in verdict-neutral specimen, generated
  experiment index, adversarial tests, and a runnable offline example.
• Add a generated CLI/API reference page and presentation gates that reject stale
  reference or experiment-index content.
• Clarify VSTD's role as a verification-domain language and interchange layer that
  preserves, rather than replaces or strengthens, native verifier results.
• Add the experimental SCITT adapter, rerunnable real-COSE specimen with ephemeral keys, explicit semantic
  boundary, and adversarial composition tests without claiming IETF review or payload
  truth from registration.
• Surface zero-identity/zero-knowledge (ZIZK) artifact-first TRUST as governing
  architecture, publish the bounded RISC Zero reference mechanism and exact recorded
  public proof artifacts, and keep only unfinished mechanisms experimental while
  preserving unresolved horizons and native-system authority.
• Bind the recorded RISC Zero proof to the image produced from the tracked guest and
  locked toolchain in the governed offline verifier, rather than accepting source/proof
  correspondence from a neighboring historical image identifier.
• Formally distinguish TRUST as mechanism-earned forward artifact support, ROT as typed
  time-indexed degradation of current admissibility, and RUST as an inverse-TRUST memetic
  causal backtrace toward recorded ancestor states. None is actor-tied trust or a scalar;
  reachability alone never infers guilt, responsibility, or causal localization.
• Present current reports, schemas, module descriptions, and examples under the full
  VSTD-1 and VSTD-2 numbered-profile identifiers; remove retired partial-profile object identifiers from
  active readers and add a regression preventing their return.
• Add normative artifact-control mechanism version 1 with exact-byte file/directory
  freezing, SHA-256 plus SHA3-256 artifact-derived identities, observable read-only
  guards, readable finite self-closing Ed25519 seals, external anchor checks, and
  copy-on-write thaw descendants. Sealing is not encryption and supplies no actor trust,
  semantic correctness, trusted time, or numbered VSTD profile result.
• Document multi-temporal realms, discrete and continuous coexistence, causal and
  problem-space partial orders, atemporal versus temporal capsules, explicit cross-realm
  mappings, and future constrained language-model transition verification without
  claiming continuous mediation, inference-law implementation, or textual truth.

Claim boundaries and validation

• Require thawed_artifact_status and vstd artifact status to verify an actual supplied,
  cleanly sealed parent and every recorded parent coordinate before returning
  THAWED_CLEAN or THAWED_DIRTY. Sidecar-only agreement is now NOT_ESTABLISHED; even a
  verified current match does not authenticate the historical copy operation or external
  parent continuity.
• Preserve final filesystem-entry identity during artifact creation: freeze refuses
  symbolic-link sources, and bundle, thaw-descendant, and sidecar outputs refuse every
  preexisting lexical entry, including dangling symbolic links, without claiming universal
  race-free filesystem security.
• Require authoritative freeze-manifest, payload, seals-container, and seal-envelope
  members to have ordinary lexical types; linked external or in-bundle targets cannot lend
  bytes to bundle closure, while verified outer read aliases and ordinary hard-link
  byte-and-path semantics remain explicitly distinct.
• Remove the live SimulacraBench rehearsal and its front-door promotion; the repository
  never contained or reproduced the submission, hosted image, hardware, or protected
  evaluation identified by that name.
• Correct generic-run wording: digest validation is an integrity check, external
  references remain unattested until dereferenced and verified, same-path output
  extraction is not independent verification, and unverified determinism is UNKNOWN.
• Publish a Pages guide index and enforce language, title, viewport, main-region, skip-link,
  image-alt, labelled-navigation, generated-reference, and local-link checks in CI.
• Preserve explicit ordered-list starting numbers in generated Pages so procedures split
  by code blocks retain their source step numbers instead of restarting at one.
• Require CodeQL security-extended Python analysis in the protected repository-check
  aggregate with only read access to content and write access to security results.
• Fail closed on malformed generic-run receipts, publish their exact schema, and dispatch
  VSTD-1 by its required receipt profile.
• Package every normative specification, verify byte identity, and smoke-test the built
  wheel outside the source checkout so installed specification bindings cannot silently
  become unavailable.
• Bind the bundled checker to VSTD-1, record actor and execution separation explicitly,
  and never infer independent actors from a historical field name, repeated runs, or
  matching results.
• Reject self-promoted independence even when every supplied status and digest agrees;
  the generic-run compatibility path never derives EVIDENCED from serialized references.
  The distinct VSTD-5 path reruns all seven separation propositions and does not upgrade
  the legacy generic-run fields.
• Require the real optional SCITT/COSE cryptographic example in the protected
  repository-check aggregate
  rather than allowing its dependency-gated tests to disappear from the base matrix.
• Close generic-run control structures while retaining the released refutation-extension
  map, make common receipt commands honor --json, and lock validate as an
  integrity/profile check rather than a claim verifier.

Graph and conformance semantics

• Add a zero-dependency evidence execution core that resolves and rehashes exact evidence
  bytes, pins a registered mechanism implementation digest, enforces byte/item bounds,
  reruns the mechanism, and preserves PASS, FAIL, or UNKNOWN under explicit trust roots.

• Add an evidence-bound VSTD-4 path and replayable receipt form. Compatibility
  vstd4_depth remains a NOT_ESTABLISHED candidate; only exact passing VSTD-1/2/3 and
  fourteen-rung mechanisms plus an accepted kernel witness admit VSTD-5.

• Implement the VSTD-5 reference mechanism and receipt: seven evidence-bound separation
  dimensions, duplicate-witness/evidence refusal, exact admitted-certificate and
  corroboration binding, typed binding/identity/separation/corroboration errors,
  disagreement preservation, embedded evidence, and offline result recheck. Independence
  fails closed on any identity or separation defect without parsing error-message text.
  Witness identities and assertions serialize separately and in order, so duplicate,
  orphan, missing, and reused-identity error inputs remain replayable instead of collapsing
  during receipt construction. Keep permissive malformed-input assessment distinct from
  portable receipt admission: the builder now raises unless the strict schema and complete
  verdict-material evidence coverage hold, and the rechecker applies the same zero-dependency
  gate before replay. The rechecker also compares the complete carried VSTD-4 entry, and
  corroboration_class is mechanism-bound rather than relabelable metadata. This does not
  claim a real external witness or independent implementation; a positive observation with
  unresolved independence is overall UNKNOWN.

• Add evidence-bound Graph profile computation and replay. The compatibility graph_level
  path remains caller-supplied; the new path reruns every member, ancestor, and reached-edge
  rating mechanism bound to the exact Graph, members, collection, and claim before profile
  1–5 can report ESTABLISHED. Profile zero remains NOT_ESTABLISHED.

• Add VSTD-GRAPH-ASSURANCE-1 and AssuranceLedger for hash-chained edge-local TRUST, ROT, RUST,
  challenge-ledger projection, additive conflict declaration/resolution, structural RUST concentration,
  explicit causal localization, and bounded artifact-relative BLAME/GUILT propositions.
  Each TRUST event binds one exact transformation, its inputs/output, the historical Graph,
  and prerequisite TRUST events; current eligibility recursively fails closed when any bound
  dependency degrades or conflicts. Duplicate paths remain set-valued, historical graph bytes
  remain immutable, and topology alone earns no causal or moral conclusion. BLAME establishes
  bounded responsibility or material contribution. GUILT is not BLAME in the opposite
  direction: it composes separately bound responsibility, exact scoped-obligation
  applicability, and same-obligation violation components, then binds their exact event
  digests. One compound mechanism may emit all three component evaluations in one invocation;
  an opaque combined pass or decorative obligation string remains NOT_ESTABLISHED.
  Localization binds one exact passing RUST event and descendant-deviation proposition.
  Neither result establishes actor morality, reputation, automatic legal liability,
  innocence, exoneration, obligation satisfaction, or absence of hidden contributors.
  Status-conflict resolution projects the
  selected state into current admissibility; arbitrary resolved predicates remain blocked.
  The current runtime has no general non-status admissibility-effect mechanism. RUST follows
  historically recorded contributing ancestry even when current lifecycle state excludes a
  route from TRUST. New construction, evidence-bound Graph establishment, and assurance
  propagation require globally disjoint artifact/transformation identifiers so an untyped
  subject_id cannot ambiguously name both; the frozen VSTD-DATA-0.1 reader retains its
  original two namespaces.
  Add complete offline event replay,
  current TRUST filtering, and deduplicated descendant reassessment discovery.

• Preserve incompatible Graph assertions as evidence-linked conflict records and label
  rating-derived Graph profile numbers as CALLER_SUPPLIED candidates with conformance NOT_ESTABLISHED.

• Classify the current VSTD-4 candidate-depth calculation as a structural result over
  caller-supplied rung references with conformance NOT_ESTABLISHED; reject that
  candidate at the VSTD-5 entry gate even when its candidate depth is 14.

• Label compatibility Graph 2–5 candidates consistently while separately presenting the
  implemented evidence-bound reference paths. Bind complete challenge-ledger state into an
  additive current Graph view without mutating history.

Release and maintainer controls

• Mark 1.2.0 metadata as an unreleased release candidate, omit any fabricated release
  date, and require the exact tagged checkout to have TIME.md set to Status: CLEAR.
• Move the immutable-release setting check before tag creation in the documented release
  sequence and enforce it again in the tag workflow, so a disabled setting stops
  publication rather than producing a mutable release.
• Make package/reference status identify VSTD-5 as the highest exposed project
  specification with an evidence-bound reference mechanism, without claiming a real
  independent witness, and require finalized release metadata in the tag workflow.
• Publish the architecture ownership map linking normative documents, runtime validators,
  schemas, and conformance tests.
• Document the five-As human traversal over existing receipt, Graph, hardware, certificate,
  reproduction, and SCITT machinery without adding a serialized receipt format; reject duplicate Graph
  identifiers and reproduction-fidelity states inferred from declarations, matching verdicts, or
  mismatching runs.
• Restore the three non-overlapping operating controls: AGENTS.md for automated work,
  HUMANS.md for human five-As reasoning, and TIME.md for current repository
  contradictions. Development may record OPEN; the exact tagged checkout must be
  CLEAR before publication.
• Replace the developmental profile-numbered generic-run container with required neutral
  assessment_context; preserve its mechanism, bound, commitment, and refutation
  coordinates without carrying a VSTD-4 conformance field.