v0.28.0

Added

• architecture skill — Cross-Section Consistency Gate (skill manifest 0.9.0 → 1.0.0).
  Prevents downstream stages from discovering architecture gaps too late by adding
  a Cross-Section Consistency Pass inside architecture --mode design.

  • prompts/22_architecture_draft.md — adds step 2c: Cross-Section
    Consistency Pass verifying (i) every MVP operation in §23 exists in §12,
    (ii) every user-visible state in §23.3/§23.4 maps onto §14, (iii) every
    human-review action in §23.6 has a §12 contract + §14 transition + §16 audit
    event + §18 failure behaviour, (iv) every progress item in §23.4 has a §16
    observability event, (v) §24 handoffs reference only formalized or deferred
    operations. Gaps logged in §25 and §27.
  • prompts/23_quality_gate_self_check.md — adds five v0.9.0 cross-section
    consistency gates (20–24); updates Instructions to include them in the §26
    table.
  • templates/architecture_design_template.md — adds
    ### Cross-Section Consistency Gate sub-table to §26.
  • SKILL.md — adds quality gate 29.
  • manifest.json — bumped to 1.0.0.

• ux-design skill — MVP phase and testability metadata (skill manifest 0.2.0 → 0.3.0).
  Enables implementation-plan to convert E2E seeds into concrete test tasks
  without guessing and to scope MVP-0 correctly.

  • prompts/07_user_stories.md — requires phase/release-gate header on
    every story (Phase, Primary Surface, Release Gate, Depends On); updates
    validation gate.
  • prompts/10_e2e_scenario_seeds.md — requires testability metadata block
    on every seed and a Testability Summary Table in §20; updates validation gate.
  • templates/e2e-scenario-template.md — adds testability metadata fields
    to skeleton and worked example.
  • templates/user-story-template.md — adds phase + release-gate fields;
    updates Rules.
  • prompts/13_quality_gate_self_check.md — adds six gate rows, five fail
    conditions, six warning conditions.
  • SKILL.md — adds quality gate 11.
  • manifest.json — bumped to 0.3.0.

v0.27.0

Added

• architecture skill — materialize mode (skill manifest 0.8.0 → 0.9.0).
  Adds architecture --mode materialize to consolidate the base architecture
  design plus all accepted update notes into one canonical implementation-ready
  source of truth before implementation-plan.
  • prompts/materialize_02_discover_and_order.md — discovers the base
    architecture by topic slug, discovers and validates accepted update notes
    (acceptance criteria: Artifact Type = architecture_update, quality-gate
    PASS, no unresolved blocking conflicts), sorts by version/update history,
    builds a section-level patch plan, flags preliminary conflicts.
  • prompts/materialize_03_apply_patches.md — checks six BLOCKING conflict
    classes (SECTION_CONFLICT, PATCH_TARGET_MISSING, TEXT_ANCHOR_GONE,
    CONTRACT_CONFLICT, SUPERSESSION_CONFLICT, BREAKING_CHANGE_UNRESOLVED); on
    any conflict writes a materialization-blocked report and stops; otherwise
    applies patches in order, removes resolved provisional wording, updates ADRs
    / open questions / handoffs.
  • prompts/materialize_04_final_document.md — produces the 30-section
    canonical architecture (27 original + Applied Updates + Superseded Patch
    Notes + Implementation-Plan Readiness), the artifact registry, and the
    open-question ledger; runs the Materialization Quality-Gate Self-Check.
  • templates/architecture_canonical_template.md — 30-section canonical
    architecture skeleton with Applied Updates, Superseded Patch Notes, and
    Implementation-Plan Readiness sections; Artifact Type = architecture_canonical.
  • templates/artifact_registry_template.md — artifact registry skeleton
    (Current Canonical Artifacts + Superseded / Historical Artifacts + Pipeline
    Stage Summary) telling downstream agents which files to read and which are
    audit history.
  • templates/open_question_ledger_template.md — open-question ledger
    skeleton centralizing all open questions with source, owner stage, blocking
    status, and required resolution action.
  • templates/materialization_blocked_template.md — blocked report skeleton
    emitted when a BLOCKING conflict prevents safe materialization.
  • references/materialization-guide.md — full guide: when to use, what it
    owns / must not own, accepted-update discovery rules, conflict classes, output
    files, quality-gate fail/warn conditions.
  • references/patch-manifest-guide.md — Patch Manifest YAML format guide
    and update-type taxonomy (NONE / NOTE_ONLY / ADR_ONLY / CONTRACT_PATCH /
    SECURITY_PATCH / OBSERVABILITY_PATCH / STRUCTURAL_PATCH / BREAKING_CHANGE).
  • references/artifact-registry-guide.md — artifact registry purpose,
    controlled status values, artifact role vocabulary, and rules.
  • references/open-question-ledger-guide.md — ledger purpose, ID prefix
    convention, status and owner-stage values, blocking-status values.
  • Mode resolver (prompts/00_mode_resolver.md) updated: adds materialize
    to the mode table and routing JSON; adds explicit materialize triggers and
    negative triggers; updates preconditions for six modes.
  • Mode-selection guide (references/mode-selection-guide.md) updated: adds
    materialize mode description, selection logic, negative triggers, and
    downstream flow showing full pipeline through materialize to
    implementation-plan.
• architecture --mode update metadata improvements (skill manifest 0.8.0 → 0.9.0).
  Update mode output grows from 11 to 13 sections:
  • §5 Patch Manifest — machine-readable YAML consumed by materialize;
    every accepted decision maps to a patch entry with target section, operation,
    patch type, and implementation-blocking flag.
  • §12 Feedback Closure Matrix — tracks which downstream feedback items
    (ux-design Architecture Feedback, security-review findings) are resolved by
    which patches; NOT_APPLICABLE when no downstream feedback was applied.
  • Patch-Type Taxonomy added to Generation Metadata §1; multi-type allowed.
  • Skill version metadata — UNKNOWN — resolver could not determine this value replaces bare unknown; quality gate warns on any UNKNOWN field.
  • prompts/update_02_apply_decisions.md and update_03_final_document.md
    updated to instruct producing §5, §12, and classified patch types.
  • references/architecture-update-guide.md updated with taxonomy table,
    Feedback Closure Matrix guidance, and 13-section quality gate.
  • SKILL.md updated: six-mode table, references table, materialize mode
    description, method section with materialize_tasks task graph.
  • manifest.json version 0.8.0 → 0.9.0; materialize_tasks task graph and
    materialize_mandatory_gates added; mode resolver routing updated.

v0.26.0

Added

• Cross-Skill Artifact Contract across the blueprint, architecture, and
  ux-design skills (blueprint skill 0.7.0 → 0.8.0, architecture 0.7.0 → 0.8.0,
  ux-design 0.1.0 → 0.2.0). A shared standard so every generated document is both
  a human report and a machine-readable handoff artifact, preventing pipeline
  drift / artifact mismatch / unstable topic slugs as the skill graph grows.
  • references/artifact-contract.md — the canonical contract (artifact-type
    registry, stable-topic-slug rules, filename rules, required Generation
    Metadata, Source/Resolved artifacts, decision register, assumptions, open
    questions, recommended next stage, quality-gate, controlled vocabulary,
    per-skill requirements). Skills install independently (each is symlinked to
    its own dir), so the file ships identically inside each of the three
    skills; a guard test asserts the copies stay byte-identical.
  • Templates aligned (7): the blueprint, the five architecture mode
    templates (design / tech-stack / update / reconciliation / review), and the
    ux-design template now carry Artifact Type + stable Topic Slug metadata,
    an unnumbered ## Cross-Skill Artifact Contract block (Source Artifacts
    Consumed, Resolved Input Artifacts where discovery is used, and a Contract
    Field Map that points at each skill's existing decision/assumption/
    open-question/next-stage sections — alignment, not duplication), and a
    Cross-Skill Artifact Contract Gate in every self-check. No existing
    section numbering changed.
  • Prompts: the seven main generation / final-document prompts now instruct
    contract compliance with the controlled vocabulary; each SKILL.md References
    table lists artifact-contract.md.
  • Guard tests: tests/unit/test_artifact_contract.py enforces the contract
    structurally (templates include Generation Metadata / Topic Slug / Source
    Artifacts Consumed / Recommended Next Stage / Quality-Gate Self-Check; the
    five architecture mode templates include Resolved Input Artifacts; every
    self-check includes the contract gate; every skill references the contract;
    the contract file carries the registry + controlled vocabulary; the three
    copies are identical).
  • Worked examples are dated pre-contract snapshots and adopt the contract on
    next regeneration; implementation-plan / security-review / test-design
    will comply from day one.

v0.25.0

Added

• architecture skill — review / update / reconcile modes (skill
  manifest 0.6.0 → 0.7.0). The three remaining modes are now fully implemented
  (previously recognized-but-deferred), completing the architecture lifecycle:
  design → stack → update → ux-design → reconcile → review → implementation-plan.
  • Shared artifact resolver (prompts/resolve_artifacts.md,
    references/artifact-discovery-guide.md). When the user passes no filenames,
    all three modes auto-discover the most relevant prior skill/mode outputs from
    the working dir / docs / design / artifacts by topic slug + candidate
    scoring (filename suffix, section markers, recency), require the architecture
    design (STOP with a clear message if missing), ASK_USER on ambiguity, and
    embed a Resolved Input Artifacts table in every output.
  • review — evaluates architecture quality without changing it: a
    10-point score breakdown with justifications, blocking/warning/polish issue
    classification, readiness assessments, and recommended next actions →
    <topic-slug>-architecture-review.md (19 sections). It is the safe default
    when bare architecture is invoked and an architecture already exists.
  • update — applies already-accepted decisions (priority source: a
    tech-stack with Architecture Update Required? = Yes, then accepted
    reconciliation, security-review, newer blueprint, explicit user decision;
    never ux-design directly) into <topic-slug>-architecture-update.md (11
    sections) — an update note that does not overwrite the design document by
    default.
  • reconcile — turns downstream feedback (primarily a ux-design
    Architecture Feedback section) into findings, a missing-architecture-support
    map, a minimal patch plan, and an explicit Architecture Update Required?
    verdict + handoff to update → <topic-slug>-architecture-reconciliation.md
    (11 sections). It does not patch the architecture by default and does not
    blindly accept a downstream artifact that contradicts the blueprint.
  • No-silent-mutation rule across all three modes; three new task graphs
    (review_tasks / update_tasks / reconcile_tasks) + mandatory gates; the
    mode resolver and mode-selection guide updated (bare architecture +
    existing architecture → review, never update); three templates, four
    references, three worked translation-system examples that chain off the
    existing stack (Architecture Update Required? = Yes) and ux-design
    (Architecture Feedback) examples; and Phase-5 guard tests.

Changed

• AGENTS.md: the architecture skill is now described as five implemented
  modes.

v0.24.0

Added

• ux-design skill (new; skill manifest 0.1.0). The fifth bundled skill and
  the fourth stage of the design chain: research-pipeline → blueprint → architecture → ux-design → implementation-plan. It consumes an architecture
  design document (architecture --mode design output; the matching blueprint
  and tech-stack are optional) and emits <topic-slug>-ux-design.md. Pure
  prompt-driven (no CLI/MCP backend); auto-discovered by setup.
  • User-story-driven, not screen-drawing. It separates Skill Operator UX
    (how the user drives the skill workflow) from Target Software UX (how end
    users and agents interact with the product), and produces structured user
    stories (preconditions, main / alternative / failure-recovery flows,
    user-visible states resolving to the architecture state model, acceptance
    criteria), core journeys, surface-specific UX (only for
    architecture-supported surfaces), human-in-the-loop UX, error / empty /
    loading / degraded / recovery states, acceptance criteria, and Gherkin-style
    E2E scenario seeds (seeds, not executable tests).
  • Mandatory architecture feedback. §21 records UX-exposed architecture gaps
    (missing states / events / review schema / permissions) with severity and
    recommends architecture --mode reconcile when needed — the UX→architecture
    feedback loop.
  • Input discovery + fail-fast. Auto-discovers *-architecture-design.md in
    the working dir / docs / design / artifacts, ranks candidates, and
    STOPs with a clear message if none is found (it does not run from a blueprint
    alone). Hybrid is the default operating mode.
  • 22-section + Appendix-A output template, 13-task manifest, 13 prompts, 3
    templates, 4 references, a worked translation-system example, and structural
    tests (tests/unit/test_skill_ux_design.py).

Changed

• architecture skill references updated to reflect that ux-design now
  exists (the mode-selection and Experience-Architecture guides no longer call it
  a "future skill not yet built"; no behavior change, architecture manifest stays
  0.6.0).
• AGENTS.md: "Four skills" → "Five skills" with a ux-design row and description.

v0.23.0

Changed

• architecture skill — design/stack mode split (skill manifest 0.5.0 →
  0.6.0). Phase 1 of the architecture-skill mode-split improvement plan
  (docs/architecture-skill-mode-split-improvement-plan.md). The skill becomes
  one skill with internal modes instead of mixing architecture design and
  tech-stack selection in one pass. Tech/domain-neutral; the existing
  prompts/template/example are extended, not rewritten.
  • Mode resolver. A new prompts/00_mode_resolver.md + a ## Modes section
    in SKILL.md + references/mode-selection-guide.md select the mode (design
    / stack / update / review / reconcile) from an explicit mode
    argument or automatic detection. design and stack are fully specified;
    update reuses design mode's existing regenerate/patch/compare/adr-only/
    resume machinery; review/reconcile are recognized but deferred (never
    silently mutate an architecture).
  • design mode consumes blueprint §9 and §19. It now reads the blueprint's
    §9 Product Experience Direction into a new §23 Experience Architecture
    section (interaction surfaces, user-visible states mapped onto the §14 state
    model, feedback/progress, error/recovery, human-review flow, trust/
    transparency, UX handoff — architecture-level UX support, not UX design), and
    reflects the §19 Recommended Next Stages routing in a new §24 Recommended
    Next Stages and Downstream Handoffs section (tech-stack / ux-design /
    security-review / test-design handoffs + update/reconciliation triggers). The
    design output grows from 25 to 27 sections.
  • Provisional-tech discipline. §7 keeps the tech stack provisional with
    new §7.1 Provisional Tech Assumptions and §7.2 Tech-Stack Selection Handoff
    sub-sections whenever §19 routes tech-stack-selection = RUN/DEFER; final
    selection is owned by stack mode. Five new §26 self-check gates enforce
    this (Product Experience Direction consumed, Experience Architecture
    produced, Recommended Next Stages consumed, tech-stack-provisional, and
    downstream-handoffs-present).
  • stack mode. A separate 6-task graph (prompts/stack_01–stack_06,
    templates/architecture_tech_stack_template.md,
    references/tech-stack-selection-guide.md) selects the concrete technology
    stack against the architecture — one decision per area with alternatives,
    rationale, risk, reversibility, and architecture impact — and ends with an
    explicit Architecture Update Required? verdict. It satisfies the
    architecture and never redesigns it; on a genuine conflict it declares the
    update instead of rewriting. Output: <topic-slug>-architecture-tech-stack.md.
  • A worked stack example (examples/translation_tech_stack_example.md) is
    added and the design example is updated to 27 sections.

v0.22.0

Changed

• blueprint skill — adaptive next-stage routing clarity (skill manifest
  0.6.0 → 0.7.0). A focused follow-up to the §19 router and §9 Product
  Experience Direction, acting on an external review of a generated blueprint.
  No redesign — the changes are columns, a couple of small tables, and a short
  rationale, all tech/domain-neutral and folded into the existing
  prompts/templates/reference/example:
  • §19.2 Depends On column. The Stage Recommendations table now states
    each stage's prerequisite explicitly (e.g. security-review and ux-design
    depend on architecture-design), distinct from the revisit trigger, so the
    pipeline no longer looks more rigid — or more parallel — than intended.
  • §19.4 Recommended Pipeline split. The single flat pipeline list is now a
    Recommended Linear Path (core ordered sequence) plus a Conditional
    Follow-up Gates table (Gate | Run When | Typical Input | Output), so
    deferred conditional stages — chiefly architecture-update /
    architecture-reconciliation — are never silently dropped from the routing
    picture.
  • §19.3 ASK_USER Decision Rationale. When no stage is ASK_USER, §19 now
    justifies why (each high-impact unknown is already answered by the source
    report / §9, deferred to architecture-design, or delegated to
    security-review, with a named owner) instead of looking over-confident;
    FAIL if ASK_USER is absent despite an unresolved high-impact unknown with
    no downstream owner.
  • §19.1 complexity score labelled a routing heuristic. The output now
    states that the / 21 score is a routing heuristic, not a formal project
    estimate, to be revisited after architecture-design — preventing false
    precision.
  • §9.4 / §9.5 interaction-mode Classification. Every interaction mode now
    carries a controlled classification — primary surface / secondary
    surface / wrapper / integration surface / future surface — with an
    explicit rule that "AI Skill" must be disambiguated (usually a
    wrapper/integration surface around the CLI/core, not a separate runtime) and
    never conflated with MCP (a tool surface for external AI agents).
  • Quality gates. The Product Experience Gate gains an "Interaction modes
    classified" row; the Adaptive Stage-Gate Recommendation Gate gains
    "Stage table has Depends On", "Linear-path vs conditional-gates split",
    "ASK_USER absence explained", and "Complexity score labelled heuristic" rows,
    plus matching WARNING conditions. The §19 output-discipline budget was
    updated so the additions (columns + small tables) do not contradict the
    "keep §19 compact" rule.

v0.21.0

Added

• blueprint skill — Recommended Next Stages (adaptive stage-gate routing;
  skill manifest 0.5.0 → 0.6.0). The blueprint is now the first adaptive
  stage-gate router after research extraction: instead of leaving every
  optional downstream stage to manual choice, it evaluates the project and
  recommends which stages should run next — with evidence, confidence, and
  revisit triggers — without silently expanding the pipeline. Folded into the
  existing prompts/templates/example (no new manifest tasks),
  tech/domain-neutral:
  • New §19 Recommended Next Stages (the output template grows from 19 to 20
    sections; §19 sits after the technical-design handoff and before the
    Traceability Appendix, which renumbers to §20). It contains a Pipeline
    Complexity Assessment (seven 0–3 dimensions — user-facing complexity,
    technical ambiguity, security/privacy risk, AI/LLM uncertainty, integration
    complexity, human-review complexity, testing/E2E importance — with a total
    / 21 and a simple/lightweight/medium/complex workflow class), a Stage
    Recommendations table covering architecture-design, tech-stack-selection,
    ux-design, security-review, test-design, architecture-update, and
    architecture-reconciliation, a short Recommended Pipeline, and a
    Stage-Gate Decision Log.
  • Controlled decision vocabulary — every stage uses exactly one of
    RUN / SKIP / DEFER / ASK_USER (no vague "maybe / consider / nice to
    have"). Every RUN/ASK_USER needs evidence, every SKIP a reason, every DEFER
    a revisit trigger; recommendations are overrideable defaults.
    architecture-design is normally RUN; architecture-update and
    architecture-reconciliation default to DEFER at blueprint stage (no
    architecture document exists yet).
  • Adaptive Stage-Gate Recommendation Gate added to the quality-gate prompt
    (Gate 9) and the Appendix A self-check (seven rows: section exists,
    controlled decisions, RUN evidence, SKIP reason, DEFER revisit trigger,
    ASK_USER missing-info, and Product Experience Direction informs the
    recommendations), with explicit FAIL and WARNING conditions.
  • Product Experience Direction integration — §9 signals drive the routing
    (human review → ux-design / test-design; external egress → security-review;
    CLI-first → ux-design DEFER; future MCP → tech-stack-selection RUN/DEFER),
    making §9 actionable rather than decorative.
  • New reference references/adaptive-stage-gate-routing.md — the decision
    vocabulary, per-stage RUN/SKIP/DEFER/ASK_USER rules, complexity scoring, the
    PED-integration table, the output-discipline budget, and the gate
    definition. SKILL.md, prompts 04/05, the template, the example, and both
    checklists were updated; the shipped example models a realistic 16/21
    "complex" routing recommendation.

Changed

• blueprint SKILL.md now documents 20 sections, adds adaptive-routing trigger
  phrases ("what should we build next?", "which design stages should we run
  next?"), and a ninth quality gate.

v0.20.0

Added

• blueprint skill — Product Experience Direction (lightweight UX-intent
  support; skill manifest 0.4.0 → 0.5.0). The blueprint now captures enough
  product-experience direction to keep downstream architecture and
  implementation from inventing UX assumptions, without turning blueprint into a
  full UX-design stage. The boundary is explicit: blueprint defines UX intent
  · architecture defines the UX-enabling technical structure · a later UX-design
  skill defines the detailed experience · implementation-plan turns it into
  build tasks. Folded into the existing prompts/templates/example (no new
  manifest tasks), tech/domain-neutral:
  • New §9 Product Experience Direction (the output template grows from 18
    to 19 sections; §9 sits after Workflow Model and before Logical
    Architecture, so UX intent exists before the architecture handoff). It
    captures, compactly (1–2 pages, tables): Primary Experience Thesis · Primary
    User / Operator · Primary Job-to-Be-Done · Primary Interaction Mode
    (CLI/Web/GUI/TUI/API/AI-Skill/MCP/Hybrid, with MVP stage + rationale) ·
    Secondary / Future Interaction Modes · Critical Trust, Control, and
    Transparency Requirements · Human-in-the-Loop Experience · Failure and
    Recovery Expectations · UX Assumptions for Architecture · Product Experience
    Handoff to Architecture.
  • Product Experience Gate added to the quality-gate prompt and the
    Appendix A self-check (eight rows: primary user, job-to-be-done, experience
    thesis, interaction mode, trust/control/transparency, human-in-the-loop,
    failure/recovery, UX handoff), with explicit FAIL and WARNING conditions.
  • UX-intent boundary enforced — §9 must not contain screen layout,
    wireframes, CSS/visual design, full user journeys, exact CLI syntax/flags,
    exact MCP/API schemas, copywriting, mobile navigation, full accessibility
    checklists, or implementation tasks (UX over-reach is a gate FAIL). SKILL.md
    gains UX trigger phrases and negative triggers (detailed UI/UX → a later
    UX-design skill); the forbidden-content checklist gains a detailed-UX
    section.
  • New reference references/product-experience-direction.md — the section
    template, the boundary rule, the grill-me-style clarification-question
    format (good vs. bad questions), the Product Experience Gate, and the output
    budget.
  • Chain handoff — the architecture skill's blueprint-parse prompt now
    extracts product_experience (interaction mode, trust/control/transparency,
    human-in-the-loop, failure/recovery, UX assumptions) so it preserves UX
    intent instead of inventing it.
  • Updates the manifest, SKILL.md, prompt 04 (generate), prompt 05
    (quality-gate), the output template, both test checklists, and the worked
    example (now modelling §9 + a PASS-with-warning PE-gate row), with new guard
    tests including a Copilot description-length check.

v0.19.5

�[38;2;0;136;0;03m## Changed�[39;00m

�[38;2;102;102;102m�[39marchitecture�[38;2;187;187;187m �[39mskill�[38;2;187;187;187m �[39m—�[38;2;187;187;187m �[39mprovenance�[38;2;187;187;187m �[39m�[38;2;102;102;102m&�[39m�[38;2;187;187;187m �[39moutput�[38;2;102;102;102m-�[39mdiscipline�[38;2;187;187;187m �[39mhardening�[38;2;187;187;187m �[39mpass�[38;2;187;187;187m �[39m(skill�[38;2;187;187;187m �[39mmanifest�[38;2;187;187;187m �[39m�[38;2;102;102;102m0.4�[39m�[38;2;102;102;102m.0�[39m�[38;2;187;187;187m �[39m→�[38;2;187;187;187m �[39m�[38;2;102;102;102m0.5�[39m�[38;2;102;102;102m.0�[39m).�[38;2;102;102;102m�[39m�[38;2;187;187;187m �[39mFourth�[38;2;187;187;187m �[39mreview�[38;2;102;102;102m-�[39mdriven�[38;2;187;187;187m �[39mpass.�[38;2;187;187;187m �[39mThe�[38;2;187;187;187m �[39mreviewed�[38;2;187;187;187m �[39mdocument�[38;2;187;187;187m �[39mwas�[38;2;187;187;187m �[39mconfirmed�[38;2;187;187;187m �[39m�[38;2;170;34;255;01mgenerated�[39;00m�[38;2;187;187;187m �[39m�[38;2;170;34;255;01mby�[39;00m�[38;2;187;187;187m �[39mv0�[38;2;102;102;102m.19�[39m�[38;2;102;102;102m.4�[39m�[38;2;187;187;187m �[39m(metadata�[38;2;187;187;187m �[39mreports�[38;2;187;187;187m �[39mskill�[38;2;187;187;187m �[39m�[38;2;102;102;102m0.4�[39m�[38;2;102;102;102m.0�[39m)�[38;2;187;187;187m �[39m�[38;2;170;34;255;01mand�[39;00m�[38;2;187;187;187m �[39mthe�[38;2;187;187;187m �[39mv0�[38;2;102;102;102m.19�[39m�[38;2;102;102;102m.4�[39m�[38;2;187;187;187m �[39mfixes�[38;2;187;187;187m �[39mverified�[38;2;187;187;187m �[39m�[38;2;170;34;255;01mas�[39;00m�[38;2;187;187;187m �[39mlanded�[38;2;187;187;187m �[39m(�[38;2;170;34;255;01mdata�[39;00m�[38;2;102;102;102m-�[39megress�[38;2;187;187;187m �[39m�[38;2;102;102;102m+�[39m�[38;2;187;187;187m �[39mverification�[38;2;102;102;102m-�[39m�[38;2;170;34;255;01mtable�[39;00m�[38;2;187;187;187m �[39m§17�[38;2;102;102;102m.12�[39m,�[38;2;187;187;187m �[39mzero�[38;2;187;187;187m �[39munchecked�[38;2;187;187;187m �[39mcheckboxes).�[38;2;187;187;187m �[39mFive�[38;2;187;187;187m �[39mtargeted�[38;2;187;187;187m �[39madditions,�[38;2;187;187;187m �[39mfolded�[38;2;187;187;187m �[39m�[38;2;170;34;255;01minto�[39;00m�[38;2;187;187;187m �[39mexisting�[38;2;187;187;187m �[39mprompts�[38;2;187;187;187m �[39m�[38;2;102;102;102m+�[39m�[38;2;187;187;187m �[39mself�[38;2;102;102;102m-�[39m�[38;2;170;34;255;01mcheck�[39;00m�[38;2;187;187;187m �[39m(�[38;2;170;34;255;01mno�[39;00m�[38;2;187;187;187m �[39m�[38;2;170;34;255;01mnew�[39;00m�[38;2;187;187;187m �[39mmanifest�[38;2;187;187;187m �[39mtasks),�[38;2;187;187;187m �[39mtech�[38;2;102;102;102m-�[39mneutral�[38;2;102;102;102m:�[39m

�[38;2;102;102;102m-�[39m�[38;2;187;187;187m �[39m�[38;2;102;102;102m�[39mDecision�[38;2;187;187;187m �[39mevidence�[38;2;187;187;187m �[39m�[38;2;102;102;102m/�[39m�[38;2;187;187;187m �[39mprovenance�[38;2;102;102;102m�[39m�[38;2;187;187;187m �[39m—�[38;2;187;187;187m �[39mhigh�[38;2;102;102;102m-�[39mimpact�[38;2;187;187;187m �[39m§3�[38;2;187;187;187m �[39mdecisions�[38;2;187;187;187m �[39mcarry�[38;2;187;187;187m �[39ma�[38;2;187;187;187m �[39mDecision�[38;2;187;187;187m �[39mEvidence�[38;2;187;187;187m �[39m�[38;2;170;34;255;01mvalue�[39;00m�[38;2;187;187;187m �[39m�[38;2;170;34;255;01mand�[39;00m�[38;2;187;187;187m �[39mare�[38;2;187;187;187m �[39m�[38;2;170;34;255;01mnever�[39;00m�[38;2;187;187;187m �[39mlabelled�[38;2;187;187;187m �[39muser-confirmed�[38;2;187;187;187m �[39m�[38;2;170;34;255;01mwithout�[39;00m�[38;2;187;187;187m �[39man�[38;2;187;187;187m �[39mexplicit�[38;2;187;187;187m �[39m�[38;2;170;34;255;01msource�[39;00m;�[38;2;187;187;187m �[39munclear�[38;2;187;187;187m �[39mprovenance�[38;2;187;187;187m �[39mdowngrades�[38;2;187;187;187m �[39m�[38;2;170;34;255;01mto�[39;00m�[38;2;187;187;187m �[39marchitecture_assumption�[38;2;187;187;187m �[39m�[38;2;102;102;102m+�[39m�[38;2;187;187;187m �[39mreview.
�[38;2;102;102;102m-�[39m�[38;2;187;187;187m �[39m�[38;2;102;102;102m�[39mRaw�[38;2;187;187;187m �[39m�[38;2;170;34;255;01msource�[39;00m�[38;2;187;187;187m �[39mcontent�[38;2;187;187;187m �[39mforbidden�[38;2;187;187;187m �[39m�[38;2;170;34;255;01min�[39;00m�[38;2;187;187;187m �[39m�[38;2;170;34;255;01mlogs�[39;00m�[38;2;187;187;187m �[39m�[38;2;170;34;255;01mby�[39;00m�[38;2;187;187;187m �[39m�[38;2;170;34;255;01mdefault�[39;00m�[38;2;102;102;102m�[39m�[38;2;187;187;187m �[39m�[38;2;170;34;255;01mfor�[39;00m�[38;2;187;187;187m �[39mexternal�[38;2;102;102;102m-�[39mmodel�[38;2;187;187;187m �[39msystems�[38;2;187;187;187m �[39m—�[38;2;187;187;187m �[39mredaction�[38;2;187;187;187m �[39m�[38;2;102;102;102m+�[39m�[38;2;187;187;187m �[39mlog�[38;2;102;102;102m-�[39m�[38;2;170;34;255;01msnapshot�[39;00m�[38;2;187;187;187m �[39m�[38;2;102;102;102m+�[39m�[38;2;187;187;187m �[39mprovider�[38;2;102;102;102m-�[39m�[38;2;170;34;255;01mwrapper�[39;00m�[38;2;187;187;187m �[39mtests�[38;2;187;187;187m �[39m�[38;2;170;34;255;01mas�[39;00m�[38;2;187;187;187m �[39ma�[38;2;187;187;187m �[39m�[38;2;170;34;255;01mrelease�[39;00m�[38;2;102;102;102m-�[39mblocking�[38;2;187;187;187m �[39m§17�[38;2;102;102;102m.12�[39m�[38;2;187;187;187m �[39mgate�[38;2;187;187;187m �[39m(operator�[38;2;187;187;187m �[39mconfig�[38;2;187;187;187m �[39malone�[38;2;187;187;187m �[39minsufficient).
�[38;2;102;102;102m-�[39m�[38;2;187;187;187m �[39m�[38;2;102;102;102m�[39mWarning�[38;2;187;187;187m �[39msurfacing�[38;2;102;102;102m�[39m�[38;2;187;187;187m �[39m—�[38;2;187;187;187m �[39m�[38;2;170;34;255;01mevery�[39;00m�[38;2;187;187;187m �[39m§24�[38;2;187;187;187m �[39mWARNING�[38;2;187;187;187m �[39m�[38;2;102;102;102m/�[39m�[38;2;187;187;187m �[39mPASS�[38;2;102;102;102m-�[39m�[38;2;170;34;255;01mwith�[39;00m�[38;2;102;102;102m-�[39mwarning�[38;2;187;187;187m �[39m�[38;2;170;34;255;01mrow�[39;00m�[38;2;187;187;187m �[39m�[38;2;170;34;255;01mis�[39;00m�[38;2;187;187;187m �[39mechoed�[38;2;187;187;187m �[39m�[38;2;170;34;255;01minto�[39;00m�[38;2;187;187;187m �[39m§1�[38;2;187;187;187m �[39m(�[38;2;170;34;255;01mnew�[39;00m�[38;2;187;187;187m �[39m§1�[38;2;102;102;102m.2�[39m)�[38;2;187;187;187m �[39m�[38;2;170;34;255;01mand�[39;00m�[38;2;187;187;187m �[39m§25�[38;2;187;187;187m �[39m�[38;2;170;34;255;01mwith�[39;00m�[38;2;187;187;187m �[39mits�[38;2;187;187;187m �[39mrequired�[38;2;187;187;187m �[39m�[38;2;170;34;255;01maction�[39;00m�[38;2;187;187;187m �[39m�[38;2;170;34;255;01mand�[39;00m�[38;2;187;187;187m �[39mblocking�[38;2;187;187;187m �[39m�[38;2;170;34;255;01mstatus�[39;00m.
�[38;2;102;102;102m-�[39m�[38;2;187;187;187m �[39m�[38;2;102;102;102m�[39mStricter�[38;2;187;187;187m �[39mstandard�[38;2;102;102;102m-�[39moutput�[38;2;187;187;187m �[39mbudget�[38;2;102;102;102m�[39m�[38;2;187;187;187m �[39m�[38;2;170;34;255;01mwith�[39;00m�[38;2;187;187;187m �[39mconcrete�[38;2;187;187;187m �[39mtargets.
�[38;2;102;102;102m-�[39m�[38;2;187;187;187m �[39m�[38;2;102;102;102m�[39mBuild�[38;2;102;102;102m-�[39msequencing�[38;2;187;187;187m �[39mcap�[38;2;102;102;102m�[39m�[38;2;187;187;187m �[39m—�[38;2;187;187;187m �[39m§25�[38;2;187;187;187m �[39m≤�[38;2;187;187;187m �[39m�[38;2;102;102;102m5�[39m�[38;2;187;187;187m �[39mhigh�[38;2;102;102;102m-�[39m�[38;2;170;34;255;01mlevel�[39;00m�[38;2;187;187;187m �[39mconstraints;�[38;2;187;187;187m �[39m�[38;2;170;34;255;01mno�[39;00m�[38;2;187;187;187m �[39m�[38;2;170;34;255;01mfile�[39;00m�[38;2;102;102;102m-�[39m�[38;2;170;34;255;01mby�[39;00m�[38;2;102;102;102m-�[39m�[38;2;170;34;255;01mfile�[39;00m�[38;2;102;102;102m/�[39mtickets�[38;2;102;102;102m/�[39mPR�[38;2;102;102;102m/�[39mmigration�[38;2;187;187;187m �[39mordering.

The�[38;2;187;187;187m �[39mreview�[38;2;187;68;68m'�[39m�[38;2;187;68;68ms lower-priority items were already shipped in v0.19.3/v0.19.4 and were not re-implemented. The worked example now models a realistic PASS-with-warning and surfaces it. See CHANGELOG.�[39m