create-yss-spec 2.2.10 → 3.0.1

package/template/.codex/skills/product-design/skills/research/scripts/validate-research-package.mjs

@@ -0,0 +1,149 @@
+#!/usr/bin/env node
+import { readFileSync } from "node:fs";
+import path from "node:path";
+
+const SOURCE_LEVELS = ["primary", "direct-experience", "near-primary", "secondary", "lead-only"];
+
+function fail(message) {
+  throw new TypeError(message);
+}
+
+function requireString(value, label) {
+  if (typeof value !== "string" || !value.trim()) fail(`${label} must be a non-empty string`);
+}
+
+function requireArray(value, label, { nonEmpty = false } = {}) {
+  if (!Array.isArray(value) || (nonEmpty && value.length === 0)) fail(`${label} must be ${nonEmpty ? "a non-empty" : "an"} array`);
+}
+
+function uniqueIds(items, label) {
+  const ids = new Set();
+  items.forEach((item, index) => {
+    requireString(item?.id, `${label}[${index}].id`);
+    if (ids.has(item.id)) fail(`${label} contains duplicate id: ${item.id}`);
+    ids.add(item.id);
+  });
+  return ids;
+}
+
+function requireChoice(value, choices, label) {
+  if (!choices.includes(value)) fail(`${label} must be one of: ${choices.join(", ")}`);
+}
+
+function parseEvidence(file) {
+  let value;
+  try {
+    value = JSON.parse(readFileSync(file, "utf8"));
+  } catch (error) {
+    fail(`${path.basename(file)} must use JSON-compatible YAML 1.2 syntax: ${error.message}`);
+  }
+  if (!value || typeof value !== "object" || Array.isArray(value)) fail("evidence root must be an object");
+  return value;
+}
+
+function validate(briefFile, evidenceFile) {
+  if (!briefFile || !evidenceFile) fail("usage: validate-research-package.mjs <brief.md> <evidence.yaml>");
+  const briefPath = path.resolve(briefFile);
+  const evidencePath = path.resolve(evidenceFile);
+  if (path.dirname(briefPath) !== path.dirname(evidencePath)) fail("brief and evidence files must be adjacent");
+  if (!briefPath.endsWith("-research-brief.md") || !evidencePath.endsWith("-evidence.yaml")) fail("expected <slug>-research-brief.md and <slug>-evidence.yaml");
+  const briefSlug = path.basename(briefPath, "-research-brief.md");
+  const evidenceSlug = path.basename(evidencePath, "-evidence.yaml");
+  if (!briefSlug || briefSlug !== evidenceSlug) fail("brief and evidence files must use the same non-empty slug");
+
+  const brief = readFileSync(briefPath, "utf8");
+  const headings = ["Research Scope", "Executive Read", "Ranked UX Problems", "Source Map", "Opportunity Map", "Evidence Limitations"];
+  headings.forEach((heading) => {
+    if (!brief.includes(`## ${heading}`)) fail(`brief is missing heading: ## ${heading}`);
+  });
+
+  const evidence = parseEvidence(evidencePath);
+  if (evidence.schema_version !== 1) fail("schema_version must be 1");
+  if (evidence.mode !== "evidence-audited") fail("mode must be evidence-audited");
+  ["product", "audience", "time_horizon"].forEach((field) => requireString(evidence.scope?.[field], `scope.${field}`));
+  ["research_questions", "inclusion_criteria", "exclusion_criteria"].forEach((field) => requireArray(evidence.scope?.[field], `scope.${field}`, { nonEmpty: true }));
+
+  requireArray(evidence.search_log, "search_log", { nonEmpty: true });
+  const searchIds = uniqueIds(evidence.search_log, "search_log");
+  const searchesById = new Map(evidence.search_log.map((entry) => [entry.id, entry]));
+  evidence.search_log.forEach((entry, index) => {
+    ["channel", "query_or_corpus", "searched_at"].forEach((field) => requireString(entry[field], `search_log[${index}].${field}`));
+    requireChoice(entry.result, ["results-found", "none-found", "access-failed", "excluded"], `search_log[${index}].result`);
+  });
+
+  requireArray(evidence.evidence_items, "evidence_items", { nonEmpty: true });
+  const evidenceIds = uniqueIds(evidence.evidence_items, "evidence_items");
+  const evidenceById = new Map(evidence.evidence_items.map((item) => [item.id, item]));
+  evidence.evidence_items.forEach((item, index) => {
+    requireChoice(item.source_level, SOURCE_LEVELS, `evidence_items[${index}].source_level`);
+    requireChoice(item.visibility, ["public", "internal"], `evidence_items[${index}].visibility`);
+    requireChoice(item.stance, ["support", "counter"], `evidence_items[${index}].stance`);
+    ["source_class", "source_ref", "locator", "observed_at", "observation", "freshness"].forEach((field) => requireString(item[field], `evidence_items[${index}].${field}`));
+    requireArray(item.limitations, `evidence_items[${index}].limitations`);
+  });
+
+  requireArray(evidence.claims, "claims", { nonEmpty: true });
+  const claimIds = uniqueIds(evidence.claims, "claims");
+  const highKinds = new Map();
+  const lowerSeverityPrimary = new Map();
+  evidence.claims.forEach((claim, index) => {
+    ["problem_id", "statement"].forEach((field) => requireString(claim[field], `claims[${index}].${field}`));
+    requireChoice(claim.claim_kind, ["problem-exists", "impact", "frequency", "recommended-move"], `claims[${index}].claim_kind`);
+    requireChoice(claim.severity, ["low", "medium", "high", "critical"], `claims[${index}].severity`);
+    requireChoice(claim.frequency_signal, ["measured", "repeated-independent", "anecdotal", "unknown"], `claims[${index}].frequency_signal`);
+    requireChoice(claim.confidence, ["low", "medium", "high"], `claims[${index}].confidence`);
+    requireChoice(claim.audit_status, ["supported", "partially-supported", "unsupported", "not-audited"], `claims[${index}].audit_status`);
+    requireChoice(claim.disposition, ["publish", "qualify", "needs-deeper-research"], `claims[${index}].disposition`);
+    requireArray(claim.evidence_refs, `claims[${index}].evidence_refs`, { nonEmpty: true });
+    requireArray(claim.counter_signal_refs, `claims[${index}].counter_signal_refs`, { nonEmpty: true });
+    const supporting = claim.evidence_refs.map((ref) => {
+      if (!evidenceIds.has(ref)) fail(`${claim.id} has unresolved evidence ref: ${ref}`);
+      const item = evidenceById.get(ref);
+      if (item.stance !== "support") fail(`${claim.id} support ref is not marked support: ${ref}`);
+      if (item.source_level === "lead-only") fail(`${claim.id} uses lead-only evidence as support: ${ref}`);
+      return item;
+    });
+    claim.counter_signal_refs.forEach((ref) => {
+      if (!evidenceIds.has(ref) && !searchIds.has(ref)) fail(`${claim.id} has unresolved counter-signal ref: ${ref}`);
+      if (evidenceIds.has(ref) && evidenceById.get(ref).stance !== "counter") fail(`${claim.id} counter-signal evidence is not marked counter: ${ref}`);
+      if (searchIds.has(ref) && searchesById.get(ref).result !== "none-found") fail(`${claim.id} counter-signal search must have result none-found: ${ref}`);
+    });
+    if (claim.audit_status === "unsupported" && claim.disposition !== "needs-deeper-research") fail(`${claim.id} is unsupported and must use needs-deeper-research`);
+    if (claim.audit_status === "partially-supported" && claim.disposition !== "qualify") fail(`${claim.id} is partially supported and must use qualify`);
+    if (["high", "critical"].includes(claim.severity)) {
+      const kinds = highKinds.get(claim.problem_id) ?? new Set();
+      kinds.add(claim.claim_kind);
+      highKinds.set(claim.problem_id, kinds);
+    } else {
+      lowerSeverityPrimary.set(claim.problem_id, (lowerSeverityPrimary.get(claim.problem_id) ?? false) || supporting.some((item) => item.source_level === "primary"));
+    }
+    if (!brief.includes(claim.id)) fail(`brief does not reference key claim: ${claim.id}`);
+  });
+
+  const requiredHighKinds = ["problem-exists", "impact", "frequency", "recommended-move"];
+  highKinds.forEach((kinds, problemId) => {
+    const missing = requiredHighKinds.filter((kind) => !kinds.has(kind));
+    if (missing.length) fail(`${problemId} is high severity but is missing audited claim kinds: ${missing.join(", ")}`);
+  });
+  lowerSeverityPrimary.forEach((hasPrimary, problemId) => {
+    if (!hasPrimary) fail(`${problemId} is lower severity but has no primary supporting evidence`);
+  });
+
+  requireArray(evidence.source_gaps, "source_gaps");
+  if (evidence.audit_summary?.status !== "complete") fail("audit_summary.status must be complete");
+  requireArray(evidence.audit_summary.audited_claim_ids, "audit_summary.audited_claim_ids", { nonEmpty: true });
+  const audited = new Set(evidence.audit_summary.audited_claim_ids);
+  audited.forEach((id) => { if (!claimIds.has(id)) fail(`audit_summary has unknown claim: ${id}`); });
+  claimIds.forEach((id) => { if (!audited.has(id)) fail(`audit_summary is missing claim: ${id}`); });
+  evidence.claims.forEach((claim) => { if (claim.audit_status === "not-audited") fail(`${claim.id} remains not-audited`); });
+  requireArray(evidence.audit_summary.notes, "audit_summary.notes");
+  return { claims: claimIds.size, evidence_items: evidenceIds.size, searches: searchIds.size };
+}
+
+try {
+  const result = validate(process.argv[2], process.argv[3]);
+  process.stdout.write(`research package valid (${result.claims} claims, ${result.evidence_items} evidence items, ${result.searches} searches)\n`);
+} catch (error) {
+  process.stderr.write(`research package invalid: ${error.message}\n`);
+  process.exitCode = 1;
+}

package/template/.codex/skills/product-design/skills/research/tests/run-scenarios.mjs

@@ -0,0 +1,72 @@
+#!/usr/bin/env node
+import { mkdtempSync, readFileSync, rmSync, writeFileSync } from "node:fs";
+import { tmpdir } from "node:os";
+import path from "node:path";
+import { spawnSync } from "node:child_process";
+import { fileURLToPath } from "node:url";
+
+const root = path.resolve(path.dirname(fileURLToPath(import.meta.url)), "..");
+const validator = path.join(root, "scripts/validate-research-package.mjs");
+const template = JSON.parse(readFileSync(path.join(root, "assets/evidence-template.yaml"), "utf8"));
+const directory = mkdtempSync(path.join(tmpdir(), "product-design-research-"));
+const brief = path.join(directory, "sample-research-brief.md");
+const evidence = path.join(directory, "sample-evidence.yaml");
+const requiredHeadings = ["Research Scope", "Executive Read", "Ranked UX Problems", "Source Map", "Opportunity Map", "Evidence Limitations"];
+
+function run(payload, expectedStatus, expectedText) {
+  writeFileSync(brief, `${requiredHeadings.map((heading) => `## ${heading}`).join("\n\n")}\n\n${payload.claims.map((claim) => claim.id).join(" ")}\n`);
+  writeFileSync(evidence, `${JSON.stringify(payload, null, 2)}\n`);
+  const result = spawnSync(process.execPath, [validator, brief, evidence], { encoding: "utf8" });
+  if (result.status !== expectedStatus || !`${result.stdout}${result.stderr}`.includes(expectedText)) {
+    throw new Error(`expected status ${expectedStatus} containing ${expectedText}, got ${result.status}: ${result.stdout}${result.stderr}`);
+  }
+}
+
+try {
+  template.scope.product = "Example";
+  template.scope.audience = "Administrators";
+  template.scope.time_horizon = "Last 90 days";
+  template.scope.research_questions = ["Where does onboarding fail?"];
+  template.scope.inclusion_criteria = ["Direct experience in scope"];
+  template.scope.exclusion_criteria = ["Unverifiable reposts"];
+  template.search_log[0].channel = "GitHub Issues";
+  template.search_log[0].query_or_corpus = "onboarding failure";
+  template.search_log[0].searched_at = "2026-09-03";
+  template.search_log[1].channel = "GitHub Issues";
+  template.search_log[1].query_or_corpus = "onboarding completed successfully";
+  template.search_log[1].searched_at = "2026-09-03";
+  template.evidence_items[0].source_class = "issue";
+  template.evidence_items[0].source_ref = "https://example.com/issues/1";
+  template.evidence_items[0].locator = "comment-1";
+  template.evidence_items[0].observed_at = "2026-09-03";
+  template.evidence_items[0].evidence_date = "2026-09-01";
+  template.evidence_items[0].observation = "A user reports that setup stopped before completion.";
+  template.claims[0].severity = "medium";
+  template.claims[0].statement = "At least one user encountered an onboarding blocker.";
+  run(template, 0, "research package valid");
+
+  const unsupported = structuredClone(template);
+  unsupported.claims[0].audit_status = "unsupported";
+  unsupported.claims[0].disposition = "publish";
+  run(unsupported, 1, "must use needs-deeper-research");
+
+  const unresolved = structuredClone(template);
+  unresolved.claims[0].evidence_refs = ["evidence-missing"];
+  run(unresolved, 1, "unresolved evidence ref");
+
+  const falseCounter = structuredClone(template);
+  falseCounter.claims[0].counter_signal_refs = ["search-001"];
+  run(falseCounter, 1, "counter-signal search must have result none-found");
+
+  const incompleteHighSeverity = structuredClone(template);
+  incompleteHighSeverity.claims[0].severity = "high";
+  run(incompleteHighSeverity, 1, "high severity but is missing audited claim kinds");
+
+  const lowerSeverityWithoutPrimary = structuredClone(template);
+  lowerSeverityWithoutPrimary.evidence_items[0].source_level = "secondary";
+  run(lowerSeverityWithoutPrimary, 1, "lower severity but has no primary supporting evidence");
+
+  process.stdout.write("research package scenarios passed (6)\n");
+} finally {
+  rmSync(directory, { recursive: true, force: true });
+}

package/template/.codex/skills/prototype-review/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: prototype-review
-description: Use when reviewing low-fidelity UI design, wireframes, prototype links, interaction specs, or state matrices before high-fidelity HTML prototype work, Spec calibration, OpenAPI Draft, vertical slicing, or implementation.
+description: Use when independently reviewing low-fidelity UI, interaction specs, or state matrices before choosing an H1/H2 prototype profile, calibrating Spec/OpenAPI, slicing, or implementation.
 ---
 
 # Prototype Review
@@ -17,7 +17,7 @@ Run this independent gate only when UI changes affect a primary user flow, navig
 - `docs/.scratch/<feature>/design/<feature>-interaction-spec.md` or prototype link.
 - State matrix, preferably based on `docs/design/templates/state-matrix-template.md`.
 - Existing OpenAPI Draft only if the review is checking alignment; do not require OpenAPI before product design.
-- `docs/.scratch/<feature>/verification/prototype-evidence.yaml` may be created as a pending evidence record, but CLI/browser validation belongs to the following high-fidelity verification gate.
+- `docs/.scratch/<feature>/verification/prototype-evidence.yaml` may be created as a pending schema v3 record, but档位构建与浏览器验证属于后续 `gate.prototype-verified`。
 
 ## Review Gates
 
@@ -26,6 +26,7 @@ Run this independent gate only when UI changes affect a primary user flow, navig
 | Page coverage | All primary pages, entry points, and navigation exits are named |
 | Flow coverage | Main path, cancel/back, failure, retry, and completion paths are explicit |
 | State coverage | loading, empty, error, readonly, disabled, no-permission, conflict, and dirty-form states are addressed or explicitly not applicable |
+| State transition coverage | 每个状态列出进入事件、允许转换、guard、动作和退出路径；不存在只能进入不能退出的状态 |
 | Permission coverage | Hidden vs disabled vs rejected actions are clear |
 | Data coverage | Visible fields, filters, sort, pagination, forms, tables, drawers, modals, and audit/version data are listed |
 | API implication | Request/response fields, error structure, pagination/filtering, permissions, and concurrency implications can be drafted |
@@ -36,12 +37,12 @@ Run this independent gate only when UI changes affect a primary user flow, navig
 
 ## Decision Rules
 
-- If a feature has UI impact and lacks page map, user flow, prototype/wireframe, or state matrix, block high-fidelity HTML prototype work, Spec calibration, and OpenAPI Draft.
+- If a feature has UI impact and lacks page map, user flow, prototype/wireframe, or state matrix, block prototype profile selection, Spec calibration, and OpenAPI Draft.
 - If the prototype hides business rules behind generic text such as "校验失败", require field-level errors and recovery behavior.
-- If a page shows a user action but the OpenAPI implication list lacks endpoint/non-goal mapping, block high-fidelity HTML prototype work, Spec calibration, or OpenAPI Draft.
+- If a page shows a user action but the OpenAPI implication list lacks endpoint/non-goal mapping, block prototype profile selection, Spec calibration, or OpenAPI Draft.
 - If Spec P0 scope says a user can manage or configure an object but the design only shows read-only data, block until the write path or scope downgrade is explicit.
 - If a state is intentionally out of scope, record why and who owns the decision.
-- If implementation dependencies are unclear, route to `yss-router` only after the prototype passes this review.
+- If implementation dependencies are unclear, route to `yss-implementation-contract-compiler` only after the prototype passes this review.
 
 ## Output Contract
 
@@ -62,7 +63,7 @@ Run this independent gate only when UI changes affect a primary user flow, navig
 - <requirements gaps, acceptance criteria updates, non-goals, pending decisions>
 
 ### Frontend Prototype Readiness
-- <component states, data dependencies, frontend acceptance notes>
+- <component states, data dependencies, profile triggers, frontend acceptance notes>
 
 ### Lifecycle Evidence
 - <persistent review path; blockers; `gate.prototype-reviewed` candidate result>

package/template/.codex/skills/wayfinder/SKILL.md

@@ -74,7 +74,7 @@ The answer isn't part of the body — it's recorded on resolution (see [Work thr
 
 Every ticket is either **HITL** — human in the loop, worked _with_ a human who speaks for themselves — or **AFK**, driven by the agent alone. A HITL ticket only resolves through that live exchange; the agent never stands in for the human's side of it (a grilling agent that answers its own questions has broken this).
 
-- **Research** (AFK): Reading documentation, third-party APIs, or local resources like knowledge bases to surface a fact a decision waits on. Resolved by a `/research` **subagent**. Use when knowledge outside the current working directory is required.
+- **Research** (AFK): Reading documentation, third-party APIs, or local resources like knowledge bases to surface a fact a decision waits on. Resolved by a `/yss-research` **subagent** (`/research` is a deprecated alias). Use when knowledge outside the current working directory is required.
 - **Prototype** (HITL): Raise the fidelity of the discussion by making a cheap, rough, concrete artifact to react to — an outline, a rough take, a stub, or UI/logic code via the /prototype skill. Links the prototype as an asset. Use when "how should it look" or "how should it behave" is the key question.
 - **Grilling** (HITL): Conversation. The default case. Always invoke the /grilling and /domain-modeling skills.
 - **Task** (HITL or AFK): Manual work that must happen before a _decision_ can be made — nothing to decide, prototype, or research, but the discussion is blocked until it's done. Signing up for a service so its API can be judged, provisioning access, moving data so its shape can be seen. This is the one type that _does_ rather than decides — and it earns its place by unblocking a decision, not by delivering the destination. The agent drives it alone where it can (AFK); otherwise it hands the human a precise checklist (HITL). Resolved when the work is done; the answer records what was done and any resulting facts (credentials location, new URLs, row counts) later tickets depend on.
@@ -112,7 +112,7 @@ User invokes with a loose idea.
 2. **Map the frontier.** Grill again, **breadth-first** this time: fan out across the whole space rather than deep on any one thread, surfacing the open decisions and the first steps takeable now. **If this surfaces no fog** — the way to the destination is already clear, the whole journey small enough for one session — you don't need a map. Stop and ask the user how they'd like to proceed.
 3. **Create the map** (label `wayfinder:map`): Destination and Notes filled in, Decisions-so-far empty, the fog sketched into **Not yet specified**.
 4. **Create the tickets you can specify now** as child issues of the map — then wire blocking edges in a **second pass** (issues need ids before they can reference each other). Wiring sorts them into the frontier and the blocked; everything you can't yet specify stays in the fog — the **Not yet specified** section.
-5. **Fire the research subagents.** For each `research` ticket you just created, spin up a `/research` subagent to resolve it in parallel, capturing its findings on a throwaway `research/<name>` branch with a context pointer from the ticket.
+5. **Fire the research subagents.** For each `research` ticket you just created, spin up a `/yss-research` subagent to resolve it in parallel, capturing its findings on a throwaway `research/<name>` branch with a context pointer from the ticket. `/research` remains only as a deprecated compatibility alias.
 6. Stop — charting is one session's work; it hand-resolves nothing.
 
 ### Work through the map

package/template/.codex/skills/yss-antd-design/SKILL.md

@@ -1,27 +1,27 @@
 ---
 name: yss-antd-design
-description: Use when 原型设计或高保真构建需要查询 Ant Design v6 设计语言、组件 API/demo/token/semantic，或落盘 AntD CLI 证据。不用于前端代码落地（改用 yss-ui）、Vue 生产 API、生命周期批准、安装官方 antd skill，或替代 yss-prototype-stage。
+description: Use when an H2 React/AntD prototype needs versioned Ant Design v6 facts or an incremental fact-pack refresh. Not for H1, frontend implementation, lifecycle approval, or replacing yss-prototype-stage.
 ---
 
 # YSS Ant Design 事实
 
-只服务原型设计与高保真构建。把官方 `@ant-design/cli`、`design.md`、MCP 与 LLMs.txt 收成可追溯的事实引擎。阶段合同仍是 `yss-prototype-stage`。前端代码落地一律改用 `yss-ui`，不要继续调用本技能。
+只服务 H2 `flow-review` 中实际采用 React AntD 的原型，把官方 `@ant-design/cli` 与 `design.md` 收成可追溯的条件事实源。阶段合同仍是 `yss-prototype-stage`。H1 不适用；前端代码落地一律改用 `yss-ui`。
 
 ## 何时使用
 
-- 已进入 `yss-prototype-stage` 的高保真产出前后，需要 Ant Design v6 组件 / token / demo / semantic / 上游 `design.md` 事实。
-- 需要写入 `prototype-evidence.yaml` 的 AntD 段。
+- 已进入 `yss-prototype-stage` 且档位为 H2、实际采用 React AntD，需要组件 / token / demo / semantic 事实。
+- 现有 fact pack 不满足精确版本、组件覆盖、项目 Token digest 与无新 API 疑问四项 freshness 条件，需要增量补齐。
 
-不要用本技能：前端实现、Vue / Ant Design Vue 生产代码、批准门禁、生成页面流、安装官方 `antd` skill，或运行 `antd setup`。进入实现切片后停用本技能，改走 `yss-ui`。
+不要用本技能：H1、H2 非 AntD 原型、前端实现、Vue / Ant Design Vue 生产代码、批准门禁、生成页面流、安装官方 `antd` skill，或运行 `antd setup`。
 
 ## 步骤
 
-1. 检测 CLI：优先已安装的 `antd`，否则用 `npx -y @ant-design/cli`。记录实际 CLI 版本。不要在模板源执行 `npm install -g`。
+1. 先检查 `docs/design/facts/antd/<exact-version>/manifest.json`。fresh 时直接引用；不 fresh 才检测 CLI，优先已安装的 `antd`，否则用 `npx -y @ant-design/cli`。记录实际 CLI 版本，不执行全局安装。
 2. 读项目 `docs/design/design.md` 与 `docs/design/tokens/*`。
 3. 选定 `<target_antd_version>`（项目已选的 v6；未选则先查再记录，不得依赖空仓库自动检测）。此后每次知识查询都传同一版本。
 4. 设置 `ANTD_NO_AUTO_REPORT=1`。一律 `--format json`，对外说明用 `--lang zh`。
 5. 查询官方 `design.md`，再按选用组件跑 `info` / `demo` / `token` / `semantic`。命令矩阵见 `references/cli-matrix.md`。
-6. 把 JSON 路径写入证据清单。字段与 lint 适用范围见 `references/evidence.md`。
+6. 把 fact-pack manifest 或增量 JSON 路径写入证据清单。字段与 lint 适用范围见 `references/evidence.md`。
 7. 上游默认与项目 token 冲突时，以项目覆盖为准，并标记 `project_override_reviewed`。
 
 ## 硬规则
@@ -31,4 +31,5 @@ description: Use when 原型设计或高保真构建需要查询 Ant Design v6
 - 禁止把官方 `design.md` 的默认色、圆角、间距写回项目实现或覆盖项目 token。
 - 禁止在前端代码落地、垂直切片实现或 `yss-ui` 任务中调用本技能。
 - 禁止把 React props / demo / hook 当作 `yss-ui` 或 Ant Design Vue API。
+- 禁止为 H1 伪造 AntD 字段，或用 React CLI 证据宣称生产 YSS/AntDV 组件已经验证。
 - MCP 与 `llms-full.txt` 不是门禁证据。边界见 `references/boundaries.md`。

package/template/.codex/skills/yss-antd-design/references/evidence.md

@@ -1,8 +1,12 @@
 # AntD 事实证据
 
-目标文件：`docs/.scratch/<feature>/verification/prototype-evidence.yaml`，字段以 `docs/design/templates/prototype-evidence-template.yaml` 为准。本文件只说明如何填 AntD 段。
+目标文件：schema v3 `docs/.scratch/<feature>/verification/prototype-evidence.yaml`。本文件只说明 H2 实际使用 React AntD 时如何填 `profile_evidence.flow_review.prototype_library_facts`；H1 不得出现 AntD CLI 段，任何原型证据都不得冒充生产 YSS/AntDV 组件验证。
 
-## 必填
+## fact pack 优先
+
+优先引用 `docs/design/facts/antd/<exact-version>/manifest.json`。只有 exact version、组件集合和项目 Token baseline digest 相同，且不存在新 API 疑问时，`source: fact-pack` 才有效；否则用 `source: cli-run` 补充增量查询。不要为模板创建空目录或 placeholder manifest。
+
+## CLI 增量查询必填
 
 - `antd.cli_command`：实际调用的可执行文件，`antd` 或 `npx -y @ant-design/cli`。
 - `antd.cli_version`：本轮 `antd -V` 或等价输出，禁止写死模板默认值。
@@ -11,8 +15,9 @@
 - `antd.queries.components[]`：每个选用组件的 `info_ref` / `demo_ref` / `token_ref` / `semantic_ref`。
 - `design_baseline.project_design_ref`：`docs/design/design.md`。
 - `design_baseline.project_override_reviewed`：上游默认与项目 token 已对照。
-- `prototype_stack.actual_antd_version`：原型 lockfile 中的实际 React `antd@6.x`，必须与 `antd.target_antd_version` 一致。
-- `visual_semantic_mapping.components[]`：记录 v6 semantic role、项目 Token、React 原型组件、生产 YSS/Ant Design Vue 4.x 目标与 React-only API 隔离结论。
+- `actual_antd_version`：原型 lockfile 中的实际 React `antd@6.x`，必须与查询版本一致。
+- `manifest_ref` / `manifest_digest`：可复验的事实包或本轮查询清单。
+- `components_covered` 与 `project_token_baseline_digest`：支持 freshness 判断。
 
 JSON 输出存为相邻文件，证据清单只引用路径。不要把整份 `llms-full.txt` 或 MCP 对话贴进清单。
 
@@ -20,11 +25,11 @@ JSON 输出存为相邻文件，证据清单只引用路径。不要把整份 `l
 
 `antd lint` 只对可解析的 React/TSX 源码有规则命中。HTML 原型会出现空 issues（假绿）或 `skippedFiles` / `parse-error` / `partial: true`。
 
-CLI JSON 只能证明查询事实，不能单独证明原型消费了目标版本或完成了 Vue/YSS 映射。`gate.prototype-verified` 还必须通过 `yss-prototype-stage/scripts/prototype-contract.mjs` 的项目与 schema v2 证据校验、Design QA、浏览器和无障碍验证。
+CLI JSON 只能证明查询事实，不能单独证明原型消费了目标版本或完成了 Vue/YSS 映射。`gate.prototype-verified` 还必须通过 schema v3 档位证据、Design QA、浏览器和无障碍验证。
 
 | 产物 | `lint_applicable` | 做法 |
 |---|---|---|
-| `docs/.scratch/<feature>/design/prototypes/index.html` | `not-applicable` | 写 `lint_reason: html-prototype`；`lint_ref` 可空；`lint_passed` 保持 `false`，不得改成 `true` |
+| 静态 HTML / H1 | `not-applicable` | 不创建 AntD 段；由 H1 合同验证 HTML |
 | React/TSX 原型源 | `true` | 对变更文件跑 `antd lint --format json`，保存输出；无 issue 才把 `lint_passed` 设为 `true` |
 
 浏览器验证仍由 `yss-prototype-stage` 采集，不在本技能宣布 `gate.prototype-verified` 通过。

package/template/.codex/skills/yss-api-integration/SKILL.md

@@ -5,9 +5,9 @@ description: 指导在 Vue3 业务列表、表单、详情和操作 Hook 中集
 
 # API 集成 Skill
 
-本目录是 API 集成的 canonical 技能；历史名称 `api-integration` 仅通过注册表和 Router alias 解析，不再维护第二份内容。
+本目录是 API 集成的 canonical 技能；历史名称 `api-integration` 仅通过注册表和 实现合同编译器 alias 解析，不再维护第二份内容。
 
-目录、frontmatter、注册表 ID 和 Router canonical 闭包键均为 `yss-api-integration`；历史名称 `api-integration` 只在注册表和 Router alias 表中解析为本技能。
+目录、frontmatter、注册表 ID 和 实现合同编译器 canonical 闭包键均为 `yss-api-integration`；历史名称 `api-integration` 只在注册表和 实现合同编译器 alias 表中解析为本技能。
 
 ## 📋 目标
 

package/template/.codex/skills/yss-application/SKILL.md

@@ -11,7 +11,7 @@ Application 层用例编排 skill。负责协调 Domain 与 Gateway，定义事
 
 - 用户要实现 Application Service、用例编排或 `@Transactional` 边界。
 - 用户要补 App 层 MapStruct Convertor 或跨聚合协调逻辑。
-- 脚手架生成完成后进入业务实现，Router 按 `backend_impact` 加载本 skill。
+- 脚手架生成完成后进入业务实现，实现合同编译器 按 `backend_impact` 加载本 skill。
 
 ## 不适用
 
@@ -46,7 +46,7 @@ Application 层用例编排 skill。负责协调 Domain 与 Gateway，定义事
 
 - 只消费批准后的 `Slice Implementation Contract` 和当前 `work_unit`。
 - AppService 骨架可 `controlled-generation`；用例编排、事务、幂等、权限和失败行为必须 `behavior-tdd`。
-- 按 `yss-router/references/yss-skill-execution-result.md` 返回统一 `YSS Skill Execution Result`。
+- 按 `yss-implementation-contract-compiler/references/yss-skill-execution-result.md` 返回统一 `YSS Skill Execution Result`。
 - 发现新 API、权限、状态机或跨上下文影响时填入 `new_impacts` 并暂停。
 
 ## 按需读取

package/template/.codex/skills/yss-audit-log/SKILL.md

@@ -23,11 +23,11 @@ description: Use when YSS `AuditLog`, `EnableAuditLog`, audit SpEL summaries, as
 
 ## 源码索引
 
-- 源码位置不要假设固定目录；先按 `yss-source-index/references/source-location.md` 定位。
+- 源码位置不要假设固定目录；先按 `yss-skill-source-index-refresh/references/source-location.md` 定位。
 - 当前技能索引：`references/source-index.md`
 - 重点源码入口通常包括审计注解、启用注解、切面、发布服务、事件模型、默认订阅器。
 
-当组件源码变化后，用 `yss-source-index` 刷新索引；刷新或读取前先按源码定位策略确认真实位置。
+当组件源码变化后，用 `yss-skill-source-index-refresh` 刷新索引；刷新或读取前先按源码定位策略确认真实位置。
 
 ## 接入检查清单
 

package/template/.codex/skills/yss-audit-log/references/source-index.md

@@ -5,7 +5,7 @@ Source commit: `38b556a5081e6c66c0c538cbc1ef839b19996ed3`
 Source worktree: `dirty`
 Indexed source root: resolved at refresh time; set `YSS_SOURCE_ROOT` to reproduce or refresh.
 
-This file is generated by `yss-source-index/scripts/refresh-yss-skill-index.mjs`. Do not hand-edit generated sections.
+This file is generated by `yss-skill-source-index-refresh/scripts/refresh-yss-skill-index.mjs`. Do not hand-edit generated sections.
 
 Paths below are source path hints captured at refresh time. If they do not exist in the current workspace, locate sources using `YSS_SOURCE_ROOT`, CodeGraph, package/class names, Maven artifact names, or repository search before assuming the source is unavailable.
 

package/template/.codex/skills/yss-cache/SKILL.md

@@ -9,7 +9,7 @@ description: 用于 YSS 缓存框架的接入、配置、代码修改和故障
 
 ## 工作流
 
-1. 按 `yss-source-index/references/source-location.md` 定位真实源码。当前工作区有 `.codegraph/` 时先用 CodeGraph；否则读取 [source-index.md](references/source-index.md) 后用符号或 Maven 模块搜索。
+1. 按 `yss-skill-source-index-refresh/references/source-location.md` 定位真实源码。当前工作区有 `.codegraph/` 时先用 CodeGraph；否则读取 [source-index.md](references/source-index.md) 后用符号或 Maven 模块搜索。
 2. 识别使用模型：Spring Cache 单后端路由、Redis 故障降级，或 JetCache local + remote。不要混用三者的行为假设。
 3. 明确缓存名、key、TTL、更新/删除路径、跨节点一致性和序列化兼容要求。
 4. 先检查现有依赖、启动注解、配置和注解用法，再决定修改业务代码、配置还是组件。
@@ -41,4 +41,4 @@ description: 用于 YSS 缓存框架的接入、配置、代码修改和故障
 - `scripts/verify-cache-component.sh [--source-root PATH] [--consumer-root PATH --consumer-module MODULE]`：验证组件、真实 Redis、Java 8 字节码及可选消费模块。
 - `scripts/check-skill-freshness.sh [source-root]`：比较技能契约与当前组件源码；发现漂移返回 1。
 
-当组件源码变化后，用 `yss-source-index` 刷新 [source-index.md](references/source-index.md)，再运行 freshness 检查。不要手工修改生成索引。
+当组件源码变化后，用 `yss-skill-source-index-refresh` 刷新 [source-index.md](references/source-index.md)，再运行 freshness 检查。不要手工修改生成索引。

package/template/.codex/skills/yss-cache/references/source-index.md

@@ -5,7 +5,7 @@ Source commit: `38b556a5081e6c66c0c538cbc1ef839b19996ed3`
 Source worktree: `dirty`
 Indexed source root: resolved at refresh time; set `YSS_SOURCE_ROOT` to reproduce or refresh.
 
-This file is generated by `yss-source-index/scripts/refresh-yss-skill-index.mjs`. Do not hand-edit generated sections.
+This file is generated by `yss-skill-source-index-refresh/scripts/refresh-yss-skill-index.mjs`. Do not hand-edit generated sections.
 
 Paths below are source path hints captured at refresh time. If they do not exist in the current workspace, locate sources using `YSS_SOURCE_ROOT`, CodeGraph, package/class names, Maven artifact names, or repository search before assuming the source is unavailable.
 

package/template/.codex/skills/yss-cache/scripts/check-skill-freshness.sh

@@ -76,7 +76,7 @@ if [ "$annotation" -nt "$index" ] || [ "$properties" -nt "$index" ] || [ "$conne
 fi
 
 if [ "$fail" -ne 0 ]; then
-  echo "RESULT: skill/source contract drift detected; refresh with yss-source-index and review semantic references"
+  echo "RESULT: skill/source contract drift detected; refresh with yss-skill-source-index-refresh and review semantic references"
   exit 1
 fi
 

package/template/.codex/skills/yss-ddd-scaffold-generator/SKILL.md

@@ -28,7 +28,7 @@ description: 用于生成完整的 YSS DDD 多模块后端脚手架。当用户
 2. 优先运行 `node scripts/generate_and_verify_scaffold.mjs`，在同一个受控工作流中生成骨架并执行真实 Maven 验证。`generate_scaffold.mjs` 只是底层生成原语，单独返回 0 不代表脚手架完成。
 3. 检查生成的模块名、POM、机械启动入口、基础配置文件和包路径。
 4. 受控工作流必须在生成项目根目录实际执行 `./mvnw validate`、`./mvnw test` 和 `./mvnw package`；三条命令全部返回 0 后才能报告完成。
-5. 三条 Wrapper 命令通过只得到 `empty-scaffold-verified`。如需声称已满足下游技能的首切片就绪条件，必须再通过 golden first slice，得到 `first-slice-verified`。
+5. 三条 Wrapper 命令通过只得到 `empty-scaffold-verified`。如需声称已满足下游技能的首切片就绪条件，必须使用 `node scripts/run_first_slice_verification.mjs` 校验批准且版本当前的 Slice Implementation Contract、完整分层产物、skill tree digest 与根 Wrapper；只有验证器成功更新 Manifest 后才得到 `first-slice-verified`。
 
 受控验证命令由本 skill 的 `node scripts/run_scaffold_verification.mjs` 固定执行；验证器先检查 `.yss/scaffold-generation.json` 的合同元数据和 Wrapper、Java、项目级 Maven settings/profile、仓库凭据是否就绪，再在指定 evidence 目录写入每条命令的 stdout/stderr、`exit_code`、`failure_category`、耗时、执行时间和 `scaffold-verification.json`。仓库或凭据失败归为 `repository-access`，与 `project-model`、`compilation`、`bootstrap-entrypoint`、`test-failure`、`packaging` 分开；任何一条命令失败或未执行都必须阻断。
 
@@ -49,12 +49,21 @@ node scripts/generate_and_verify_scaffold.mjs \
   --contract-id <approved-scaffold-contract-id> \
   --contract-version <current-version> \
   --approval-ref <lifecycle-approval-ref> \
-  --router-draft-ref <router-draft-ref> \
+  --compiler-draft-ref <compiler-draft-ref> \
   --persisted-ref <persisted-contract-ref> \
   --contract-file /path/to/persisted-scaffold-contract.json \
   --evidence-dir /path/to/evidence/scaffold
 ```
 
+完成批准的 golden first slice 后运行：
+
+```bash
+node scripts/run_first_slice_verification.mjs \
+  --project-root /path/to/implementation-repo/my-service \
+  --slice-contract-file /path/to/approved-slice-contract.json \
+  --evidence-dir /path/to/evidence/first-slice
+```
+
 ## 生成结果应包含
 
 - 父工程 POM
@@ -75,18 +84,19 @@ node scripts/generate_and_verify_scaffold.mjs \
 - Harness 内多项目布局必须使用 `apps/backend/<project>/`；`apps/backend/` 只能是生成器的父容器，`app/backend/`、`app/frontend/` 及其子路径一律拒绝。`git-submodule` gitlink、空挂载点、detached HEAD 工作树不得覆盖生成，即使传入 `--force` 也不得当成普通目录，且不得走「请显式传入 `--force`」普通目录覆盖 / rename 路径。`--output-dir` 指向 detached HEAD 子仓时不得 mkdir、staging 或生成工程。先 `git submodule update --init` 并在子仓附加分支工作树内生成。生成器必须先调用 `gitSubmoduleScaffoldViolation`，再在 exists / `--force` / rename 之前调用 `refuseGitlinkAsRegularDirectory`。
 - 不要在 skill 里硬编码用户业务字段或真实连接信息。
 - 生成后要检查依赖关系是否仍符合分层约束。
-- 生成后必须继续使用 `yss-backend-scaffold-parent` 校验工程基线，再按垂直切片加载 `yss-domain`、`yss-application`、`yss-repository`、`yss-web-controller` 等行为 skill；各 skill 的 `references/*-layer-guide.md` 承载分层开发规范。
+- 生成时的静态依赖由技能注册表与 实现合同编译器 共同约束为 `yss-backend-scaffold-parent`、`alibaba-java-code-style`。生成后必须回到 实现合同编译器，并按批准切片加载 `yss-domain`、`yss-application`、`yss-repository`、`yss-mybatis`、`yss-web-controller`、`yss-dto`、`yss-exception`、`yss-validation`、`mapstruct`、`lombok`、`alibaba-java-code-style` 等实际命中的行为 skill。
 - 当前脚手架第一阶段仅支持经过验证的 `mysql`；未提供完整模板和验证的数据库类型不得伪装成已支持。
 - 生成后的后端工程必须使用项目根目录 `./mvnw ...` 执行构建、测试、运行和 CI 验证；不得在 README、实施记录、Ticket、Review 或 Release 中默认写裸 `mvn ...`。既有仓库确实无法使用 wrapper 时，必须记录受控例外。
 - 原型确认后，`scaffold_status=required` 才能进入本 skill；本 skill 的生成边界是工程结构、POM、配置、Wrapper 和机械模板，不是业务实现。
-- 脚手架合同必须携带 `contract_id`、`contract_version`、Router draft 引用、生命周期批准引用、持久化引用、当前版本、允许写路径、预期证据文件和验证命令；字段缺失或版本过期时阻断。
+- 脚手架合同必须携带 `contract_id`、`contract_version`、实现合同编译器 draft 引用、生命周期批准引用、持久化引用、当前版本、允许写路径、预期证据文件和验证命令；字段缺失或版本过期时阻断。
 - 新脚手架合同只接受 schema v2，必须分别携带 `project_name`、`base_package` 和 `maven_coordinates`；后者包含项目 `group_id` / `project_version`、父 POM GAV 与 `yss_components_version`，并由 CLI 原样传入。schema v1 为 `unsupported`，直接拒绝且不自动升级。
 - 合同 `profiles` 只支持 `target-domain-model`、`mybatis-plus`、`mysql`、`spring-boot-2.7-jdk8`、`javax`、`web`、`yss-internal`。普通 MyBatis、Boot 3、独立 client module、client-in-domain 和其他旧架构均为 `unsupported`，不提供回退分支。
 - 运行生成器必须传入 `--contract-file`；生成器会校验合同 `status=approved`、`current_version`、`primary_skill`、`controlled-generation`、实际输出路径和固定三条验证命令，不接受仅凭任意字符串引用的放行。
-- 生成项目必须写入 Manifest v2 `.yss/scaffold-generation.json`，记录合同 digest、Target Profile、模板 digest、下游 skill hash、generator-owned 文件 hash、严格 `generation_policy` 和完成等级；清单缺失或不一致时不得交给后续 Router。
+- 生成项目必须写入 Manifest v2 `.yss/scaffold-generation.json`，记录合同 digest、Target Profile、模板 digest、下游完整 skill tree digest、脚手架父合同与 实现合同编译器 合同 digest、generator-owned 文件 hash、严格 `generation_policy` 和完成等级；清单缺失或不一致时不得交给后续 实现合同编译器。
+- `first-slice-verified` 只能由 `run_first_slice_verification.mjs` 写入。手工改 Manifest、只生成 Controller、只通过局部模块测试或仅有结构扫描都不能升级完成等级。
 - 严禁把领域规则、状态机、权限、事务、复杂查询、错误映射、业务字段或用户可见行为塞进脚手架生成步骤。`./mvnw validate`、输出目录存在或“生成成功”都不等于生命周期批准、架构放行或 `ready-for-agent`。
 - 生命周期脚手架生成必须关闭 `--with-example`，不得把 User CRUD 或业务字段当作样板。生成器严格 `initialize-only`：非空目标、`--force`、旧项目迁移和当前模板升级一律 `unsupported`。未来若支持同一 Target Profile 内的模板升级，必须另行设计和批准，当前不预留可执行承诺。
-- 脚手架完成后，所有后续生成的后端代码必须回到 `yss-router`，消费批准且版本当前的 Slice Implementation Contract 和对应 YSS skill；业务行为使用 `behavior-tdd`，机械生成才使用 `controlled-generation`。
+- 脚手架完成后，所有后续生成的后端代码必须回到 `yss-implementation-contract-compiler`，消费批准且版本当前的 Slice Implementation Contract 和对应 YSS skill；业务行为使用 `behavior-tdd`，机械生成才使用 `controlled-generation`。
 - `.mvn/settings.xml` 只能通过 `${env.MAVEN_REPO_USERNAME}` 和 `${env.MAVEN_REPO_PASSWORD}` 读取 Maven 仓库凭据；内部仓库构建前由 CI 或本地安全环境注入变量，禁止把 Maven 仓库用户名、明文密码或 Maven 加密密码写入 skill、模板或生成工程。
 - `.mvn/maven.config` 必须显式加载项目级 `.mvn/settings.xml` 并激活 `yss-internal`；仓库 URL 来自 `YSS_MAVEN_REPOSITORY_URL`，凭据来自 `MAVEN_REPO_USERNAME` / `MAVEN_REPO_PASSWORD`。预检缺失时先于 Maven 执行阻断，日志必须脱敏。
 - Domain POM 不得依赖 YSS DTO/Exception、Web Validation、Swagger 或 Jackson；DTO/Exception/OpenAPI 注解属于 Web。MapStruct 统一 `componentModel="spring"`，父 POM 负责 processor 与 `lombok-mapstruct-binding`。
@@ -98,6 +108,7 @@ node scripts/generate_and_verify_scaffold.mjs \
 - 主脚本：`scripts/generate_scaffold.mjs`
 - 推荐的一键生成验证入口：`scripts/generate_and_verify_scaffold.mjs`
 - 受控验证器：`scripts/run_scaffold_verification.mjs`
+- 首切片验证器：`scripts/run_first_slice_verification.mjs`
 - 模板目录：`assets/templates/`
 - 分层细化参考：`references/yss-backend-scaffold-parent/` 及其子 skill
 
@@ -107,4 +118,4 @@ node scripts/generate_and_verify_scaffold.mjs \
 - 工程骨架属于 `controlled-generation`，必须记录生成器输入、预期文件和 `./mvnw ...` 编译/测试结果；不得将业务状态机、权限、事务或查询逻辑混入生成步骤。
 - 生成结果必须包含 `.yss/scaffold-generation.json`，并由验证器回勾该清单；清单与批准合同不一致时阻断。
 - `./mvnw validate`、`./mvnw test`、`./mvnw package` 必须由受控工作单元真实执行，并记录每条命令的 `exit_code`、耗时、stdout/stderr 引用和执行时间；只列命令计划、生成器成功或打印输出均视为缺失证据并阻断。
-- 生成后按统一 `YSS Skill Execution Result` 返回 changed/evidence files、实际验证结果和新增影响，再由 Router 为业务工作单元重新路由。
+- 生成后按统一 `YSS Skill Execution Result` 返回 changed/evidence files、实际验证结果和新增影响，再由 实现合同编译器 为业务工作单元重新路由。

package/template/.codex/skills/yss-ddd-scaffold-generator/agents/openai.yaml

@@ -2,4 +2,4 @@ version: 1
 interface:
   display_name: "YSS DDD Scaffold"
   short_description: "生成多模块 YSS DDD 后端脚手架与工程骨架规范"
-  default_prompt: "Use $yss-ddd-scaffold-generator only with a lifecycle-approved scaffold contract JSON to create a pure YSS DDD backend skeleton with domain, application, infrastructure, adapter, and bootstrap module boundaries; route all business code back through yss-router and the applicable YSS skills."
+  default_prompt: "Use $yss-ddd-scaffold-generator only with a lifecycle-approved scaffold contract JSON to create a pure YSS DDD backend skeleton with domain, application, infrastructure, adapter, and bootstrap module boundaries; route all business code back through yss-implementation-contract-compiler and the applicable YSS skills."

package/template/.codex/skills/yss-ddd-scaffold-generator/references/ARCHITECTURE.md

@@ -104,13 +104,13 @@ Domain Error -> Application propagation/translation -> Web Exception Translator
 
 - Scaffold 只生成机械模块、POM、Wrapper、配置、架构规则和 Manifest，不生成生产业务 CRUD。
 - 初始生成器为 `initialize-only`，非空目标失败。
-- Manifest v2 记录模板/合同/downstream skill digest 与 generator-owned 文件 hash。
+- Manifest v2 记录模板、脚手架父合同、实现合同编译器 合同、downstream 完整 skill tree digest 与 generator-owned 文件 hash。
 - 当前不支持模板升级，不得重跑初始生成器覆盖业务工程。未来如需支持同一 Target Profile 内的模板版本升级，必须另行设计、批准并验证。
 
 ## 5. 完成等级
 
 - `generated`：文件已原子生成，尚未完成 Wrapper 验证。
 - `empty-scaffold-verified`：根目录 `./mvnw validate/test/package` 全部通过。
-- `first-slice-verified`：golden first slice 的 Domain、Application、Repository/MyBatis、DTO/Web、Exception、Wire 与架构负例全部通过。
+- `first-slice-verified`：批准且版本当前的 golden first slice 已具备 Domain 模型/Gateway、Application Service 实现/事务/Query Port、Repository/Mapper/GatewayImpl/QueryAdapter、DTO/Web/Exception 与分层测试；下游完整 skill tree digest 无漂移，且根 Wrapper 全部通过。该状态只能由 `scripts/run_first_slice_verification.mjs` 写入。
 
-只有最后一级可以声明已满足下游 YSS skills 的首切片就绪条件。
+只有首切片验证器可以把 Manifest 从 `empty-scaffold-verified` 升级到最后一级；手工写值、局部测试或结构扫描均无效。只有最后一级可以声明已满足下游 YSS skills 的首切片就绪条件。