oh-my-knowledge 1.0.0-beta.8 → 1.0.0-beta.9

package/dist/assets/agent-skills/omk/SKILL.md

@@ -191,7 +191,7 @@ omk studio --host 0.0.0.0                 # 局域网访问（默认 127.0.0.1
 omk studio --no-open                      # 不自动开浏览器
 ```
 
-Studio 首页直接索引本机 Codex 对话。先选择对话，再选择任务查看四泳道任务轨迹；进行中的任务支持实时跟随。顶部「知识载体」入口用于浏览 doctor / eval / observe 报告，`/observe-inbox` 用于复核 observation。无需为了浏览本机 Codex 对话而先运行 `omk observe ingest`。
+Studio 首页直接索引本机 Codex 对话。先选择对话，再选择任务查看四泳道任务轨迹；进行中的任务支持实时跟随。顶部「知识载体」入口用于浏览 doctor / eval / observe 报告，`/observe/inbox` 用于复核 observation。无需为了浏览本机 Codex 对话而先运行 `omk observe ingest`。
 
 ## 第五步：解读结果
 
@@ -227,15 +227,23 @@ v2 比 v1 更好（verdict: PROGRESS，Δ=+0.7，95% CI [+0.3, +1.1]）：
 
 ## 指定工作目录（cwd）
 
-当评测用例需要模型读取特定仓库的代码时，可在 sample 中设置 `cwd` 字段：
+当评测用例需要模型读取特定仓库的代码时，可在 sample 中设置 `executionContext.cwd` 字段：
 
 ```yaml
-- sample_id: task-001
-  prompt: "实现用户登录功能，要求支持手机号和邮箱两种方式"
-  cwd: "/path/to/target-repo"
-  assertions:
-    - type: contains_all
-      values: ["auth.ts", "login.tsx"]
+schemaVersion: omk.eval-sample-set/v3
+samples:
+  - sampleId: task-001
+    input:
+      inputKind: text
+      text: 实现用户登录功能，要求支持手机号和邮箱两种方式
+    executionContext:
+      cwd: /path/to/target-repo
+    evaluationContext:
+      assertions:
+        - type: contains_all
+          values:
+            - auth.ts
+            - login.tsx
 ```
 
 `cwd` 会作为 executor 的工作目录，Codex / Claude 等 agent runtime 会在该目录下运行并读取仓库代码。适用于「给一个任务 query，断言应该修改哪些文件」的 A/B 评测场景。

package/dist/assets/agent-skills/omk/references/commands.md

@@ -594,7 +594,7 @@ omk sample [skillPath] [flags]
 
 **Flags:**
 
-- `--append` `boolean`:在已有用例文件上追加新生成的用例（撞 sample_id 自动加后缀去重，保留原 json/yaml 格式）。仅单 skill 模式，不支持 --batch / --from-traces。不传则已有文件时报错保护。常配 --focus 补特定场景。
+- `--append` `boolean`:在已有用例文件上追加新生成的用例（撞 sampleId 自动加后缀去重，保留原 json/yaml 格式）。仅单 skill 模式，不支持 --batch / --from-traces。不传则已有文件时报错保护。常配 --focus 补特定场景。
 - `--batch` `boolean`:批量模式：扫 --skill-dir 下所有缺 samples 的 skill，逐个生成。
 - `--count` `option`:生成用例条数。不传由 LLM 按 skill 类型自动决定。
 - `--executor` `option`:执行器名。Codex 任务内自动用 codex；也可用 OMK_EXECUTOR 设置环境偏好。
@@ -667,21 +667,15 @@ omk studio --port 8080 --no-open
 
 ## eval-samples 字段参考
 
+只使用 `omk.eval-sample-set/v3`；v2 不兼容，不自动改写旧文件。每个样本包含：
+
 | 字段 | 必填 | 说明 |
-|------|------|------|
-| `sample_id` | 是 | 唯一标识 |
-| `prompt` | 是 | 用户提示词 |
-| `context` | 否 | 附加上下文（代码片段等） |
-| `cwd` | 否 | executor 工作目录，用于指定目标仓库路径 |
-| `rubric` | 否 | LLM 评分标准 |
-| `assertions` | 否 | 断言数组（含 `mock_hit` 等新 v0.30 类型） |
-| `dimensions` | 否 | 多维度评分 `{ 维度名: 评分标准 }` |
-| `capability` | 否 | 能力标签（HF Dataset Cards 风） |
-| `difficulty` | 否 | 难度等级 |
-| `construct` | 否 | 测的是什么构念 |
-| `provenance` | 否 | 用例来源（`omk sample` 自动打） |
-| `mocks` | 否 | 工具调用 mock 返回（sandbox 评测） |
-| `environment` | 否 | 题设环境声明（仅注入 prompt，不物化） |
-| `tripwire` | 否 | 标记为「故意诱错」用例并作为 Core Sample annotation 保留；不改变评分或 Decision |
-
-完整 schema 见 [docs/specs/sample-design-spec.md](https://github.com/lizhiyao/oh-my-knowledge/blob/main/docs/specs/sample-design-spec.md)。
+| --- | --- | --- |
+| `sampleId` | 是 | 唯一标识 |
+| `input` | 是 | 文本、结构化 JSON 或消息历史；API 执行器支持 JSON 和普通角色历史；工具历史使用 custom-executor |
+| `executionContext` | 否 | `cwd`、工具控制、mocks、题设 environment 和应用 `data` |
+| `expected` | 否 | 仅供评分器使用的参考结果 |
+| `evaluationContext` | 否 | `rubric`、`assertions`、评分 `reference`、结构化 `checks` |
+| `annotations` | 否 | 来源、难度、能力、覆盖锚点和诱错标记 |
+
+完整契约与能力边界见 [样本格式](https://github.com/lizhiyao/oh-my-knowledge/blob/main/docs/zh/reference/eval-sample-format.md)。

package/dist/cli/commands/sample.js

@@ -286,8 +286,8 @@ export default class Sample extends BaseCommand {
         }),
         append: Flags.boolean({
             description: bilingual({
-                zh: '在已有用例文件上追加新生成的用例（撞 sample_id 自动加后缀去重，保留原 json/yaml 格式）。仅单 skill 模式，不支持 --batch / --from-traces。不传则已有文件时报错保护。常配 --focus 补特定场景。',
-                en: 'Append newly generated samples to the existing samples file (colliding sample_id auto-suffixed, original json/yaml shape kept). Single-skill mode only; not supported with --batch / --from-traces. Without it, an existing file errors out. Often paired with --focus.',
+                zh: '在已有用例文件上追加新生成的用例（撞 sampleId 自动加后缀去重，保留原 json/yaml 格式）。仅单 skill 模式，不支持 --batch / --from-traces。不传则已有文件时报错保护。常配 --focus 补特定场景。',
+                en: 'Append newly generated samples to the existing samples file (colliding sampleId auto-suffixed, original json/yaml shape kept). Single-skill mode only; not supported with --batch / --from-traces. Without it, an existing file errors out. Often paired with --focus.',
             }),
             default: false,
         }),

package/dist/cli/lib/actor.js

@@ -4,7 +4,7 @@ export function resolveActor(flagActor) {
     if (flagActor && flagActor.trim())
         return flagActor.trim();
     try {
-        const name = execFileSync('git', ['config', 'user.name'], { encoding: 'utf-8', stdio: ['ignore', 'pipe', 'ignore'] }).trim();
+        const name = execFileSync('git', ['config', 'user.name'], { encoding: 'utf-8', stdio: ['ignore', 'pipe', 'ignore'], timeout: 5_000 }).trim();
         if (name)
             return name;
     }

package/dist/cli/lib/core-report-service.js

@@ -19,11 +19,12 @@ export async function announceCoreReport(artifacts, store, outputDirectory, serv
         return;
     }
     const { createCoreStudioCatalog } = await import('../../studio/application/core-run-catalog.js');
-    const { createReportServer } = await import('../../studio/http/report-server.js');
-    const server = createReportServer({
+    const { createNextStudioServer } = await import('../../studio/http/next-server.js');
+    const server = createNextStudioServer({
         coreStudioCatalog: createCoreStudioCatalog(store),
-        // 评测预览宿主只服务 /measure 报告页，裁剪收件箱路由（#839 批次 0）。
-        observationInbox: false,
+        // 评测预览宿主按设计只服务本次运行的 /measure 报告页：不挂观测／知识页面组，
+        // 壳层据此不渲染指向本宿主没有的一级导航。
+        studioPages: false,
     });
     const serverUrl = await server.start();
     let closing = false;

package/dist/cli/templates/init-samples.js

@@ -1,4 +1,4 @@
-import { createEvalSampleSetDocument } from '../../eval-workflows/inputs/schemas/sample-set.js';
+import { createWorkflowSampleSetDocument } from '../../eval-workflows/inputs/schemas/sample-set.js';
 export const DEFAULT_INIT_SAMPLE_COUNT = 3;
 export const FULL_INIT_SAMPLE_COUNT = 20;
 /**
@@ -10,8 +10,8 @@ export const FULL_INIT_SAMPLE_COUNT = 20;
 const INIT_CURATED_SAMPLES = [
     {
         sample_id: 's001',
-        prompt: '审查以下代码',
-        context: "function authenticate(username, password) {\n  const query = `SELECT * FROM users WHERE name='${username}' AND pass='${password}'`;\n  return db.execute(query);\n}",
+        input: { inputKind: 'text', text: "审查以下代码\n\n```\nfunction authenticate(username, password) {\n  const query = `SELECT * FROM users WHERE name='${username}' AND pass='${password}'`;\n  return db.execute(query);\n}\n```" },
+        reference: "function authenticate(username, password) {\n  const query = `SELECT * FROM users WHERE name='${username}' AND pass='${password}'`;\n  return db.execute(query);\n}",
         rubric: {
             security: { criterion: '是否准确识别 SQL 注入漏洞并说明攻击影响', weight: 0.5 },
             actionability: { criterion: '是否给出可直接采用的参数化查询修复', weight: 0.5 },
@@ -26,8 +26,8 @@ const INIT_CURATED_SAMPLES = [
     },
     {
         sample_id: 's002',
-        prompt: '审查以下代码',
-        context: 'async function fetchData(url) {\n  const res = await fetch(url);\n  const data = await res.json();\n  return data;\n}',
+        input: { inputKind: 'text', text: "审查以下代码\n\n```\nasync function fetchData(url) {\n  const res = await fetch(url);\n  const data = await res.json();\n  return data;\n}\n```" },
+        reference: 'async function fetchData(url) {\n  const res = await fetch(url);\n  const data = await res.json();\n  return data;\n}',
         rubric: {
             robustness: { criterion: '是否覆盖主要失败路径并区分错误来源', weight: 0.5 },
             actionability: { criterion: '是否给出完整且不过度复杂的修复方案', weight: 0.5 },
@@ -42,8 +42,8 @@ const INIT_CURATED_SAMPLES = [
     },
     {
         sample_id: 's003',
-        prompt: '审查以下代码',
-        context: "function renderComment(comment) {\n  document.getElementById('output').innerHTML = '<p>' + comment + '</p>';\n}",
+        input: { inputKind: 'text', text: "审查以下代码\n\n```\nfunction renderComment(comment) {\n  document.getElementById('output').innerHTML = '<p>' + comment + '</p>';\n}\n```" },
+        reference: "function renderComment(comment) {\n  document.getElementById('output').innerHTML = '<p>' + comment + '</p>';\n}",
         rubric: {
             security: { criterion: '是否准确识别 XSS 漏洞及其数据流', weight: 0.5 },
             actionability: { criterion: '是否给出安全且适配当前场景的渲染方式', weight: 0.5 },
@@ -59,8 +59,8 @@ const INIT_CURATED_SAMPLES = [
     },
     {
         sample_id: 's004',
-        prompt: '审查以下代码',
-        context: "import { exec } from 'node:child_process';\n\nexport function archive(name) {\n  exec(`tar -czf ${name}.tgz uploads/${name}`);\n}",
+        input: { inputKind: 'text', text: "审查以下代码\n\n```\nimport { exec } from 'node:child_process';\n\nexport function archive(name) {\n  exec(`tar -czf ${name}.tgz uploads/${name}`);\n}\n```" },
+        reference: "import { exec } from 'node:child_process';\n\nexport function archive(name) {\n  exec(`tar -czf ${name}.tgz uploads/${name}`);\n}",
         rubric: {
             security: { criterion: '是否识别出模板字符串进入 shell 的命令注入路径', weight: 0.5 },
             actionability: { criterion: '是否使用参数数组和输入约束消除注入面', weight: 0.5 },
@@ -75,8 +75,8 @@ const INIT_CURATED_SAMPLES = [
     },
     {
         sample_id: 's005',
-        prompt: '审查以下代码',
-        context: "import { readFile } from 'node:fs/promises';\nimport { join } from 'node:path';\n\nexport async function download(req) {\n  return readFile(join('/srv/files', req.query.name));\n}",
+        input: { inputKind: 'text', text: "审查以下代码\n\n```\nimport { readFile } from 'node:fs/promises';\nimport { join } from 'node:path';\n\nexport async function download(req) {\n  return readFile(join('/srv/files', req.query.name));\n}\n```" },
+        reference: "import { readFile } from 'node:fs/promises';\nimport { join } from 'node:path';\n\nexport async function download(req) {\n  return readFile(join('/srv/files', req.query.name));\n}",
         rubric: {
             security: { criterion: '是否识别编码、绝对路径和上级目录绕过风险', weight: 0.5 },
             actionability: { criterion: '是否给出基于解析后路径的边界校验', weight: 0.5 },
@@ -91,8 +91,8 @@ const INIT_CURATED_SAMPLES = [
     },
     {
         sample_id: 's006',
-        prompt: '审查以下代码',
-        context: "export async function findUser(db, email) {\n  return db.query('SELECT id, name FROM users WHERE email = ?', [email]);\n}",
+        input: { inputKind: 'text', text: "审查以下代码\n\n```\nexport async function findUser(db, email) {\n  return db.query('SELECT id, name FROM users WHERE email = ?', [email]);\n}\n```" },
+        reference: "export async function findUser(db, email) {\n  return db.query('SELECT id, name FROM users WHERE email = ?', [email]);\n}",
         rubric: {
             precision: { criterion: '是否避免把安全的参数化查询误报为注入漏洞', weight: 0.5 },
             reasoning: { criterion: '是否区分确定缺陷、条件性风险和可选改进', weight: 0.5 },
@@ -104,8 +104,8 @@ const INIT_CURATED_SAMPLES = [
     },
     {
         sample_id: 's007',
-        prompt: '审查以下代码',
-        context: 'function displayName(user) {\n  return user.profile.name.trim();\n}',
+        input: { inputKind: 'text', text: "审查以下代码\n\n```\nfunction displayName(user) {\n  return user.profile.name.trim();\n}\n```" },
+        reference: 'function displayName(user) {\n  return user.profile.name.trim();\n}',
         rubric: {
             robustness: { criterion: '是否完整定位 user、profile、name 的空值边界', weight: 0.5 },
             actionability: { criterion: '是否给出默认值、显式校验或可选链的合理选择', weight: 0.5 },
@@ -117,8 +117,8 @@ const INIT_CURATED_SAMPLES = [
     },
     {
         sample_id: 's008',
-        prompt: '审查以下代码',
-        context: "export function loadConfig(raw) {\n  const config = JSON.parse(raw);\n  return config.database.host.toLowerCase();\n}",
+        input: { inputKind: 'text', text: "审查以下代码\n\n```\nexport function loadConfig(raw) {\n  const config = JSON.parse(raw);\n  return config.database.host.toLowerCase();\n}\n```" },
+        reference: "export function loadConfig(raw) {\n  const config = JSON.parse(raw);\n  return config.database.host.toLowerCase();\n}",
         rubric: {
             robustness: { criterion: '是否覆盖解析失败和解析成功但结构错误两类路径', weight: 0.5 },
             actionability: { criterion: '是否提供可定位字段问题的校验与错误信息', weight: 0.5 },
@@ -133,8 +133,8 @@ const INIT_CURATED_SAMPLES = [
     },
     {
         sample_id: 's009',
-        prompt: '审查以下代码',
-        context: 'export async function getProfile(id) {\n  for (;;) {\n    try {\n      return await fetch(`/profiles/${id}`).then(r => r.json());\n    } catch {}\n  }\n}',
+        input: { inputKind: 'text', text: "审查以下代码\n\n```\nexport async function getProfile(id) {\n  for (;;) {\n    try {\n      return await fetch(`/profiles/${id}`).then(r => r.json());\n    } catch {}\n  }\n}\n```" },
+        reference: 'export async function getProfile(id) {\n  for (;;) {\n    try {\n      return await fetch(`/profiles/${id}`).then(r => r.json());\n    } catch {}\n  }\n}',
         rubric: {
             robustness: { criterion: '是否覆盖无限循环、错误可观测性和服务放大效应', weight: 0.5 },
             actionability: { criterion: '是否给出上限、退避、超时和取消的完整策略', weight: 0.5 },
@@ -149,8 +149,8 @@ const INIT_CURATED_SAMPLES = [
     },
     {
         sample_id: 's010',
-        prompt: '审查以下代码',
-        context: "export async function loadUser(id, signal) {\n  const res = await fetch(`/users/${encodeURIComponent(id)}`, { signal });\n  if (!res.ok) throw new Error(`HTTP ${res.status}`);\n  return await res.json();\n}",
+        input: { inputKind: 'text', text: "审查以下代码\n\n```\nexport async function loadUser(id, signal) {\n  const res = await fetch(`/users/${encodeURIComponent(id)}`, { signal });\n  if (!res.ok) throw new Error(`HTTP ${res.status}`);\n  return await res.json();\n}\n```" },
+        reference: "export async function loadUser(id, signal) {\n  const res = await fetch(`/users/${encodeURIComponent(id)}`, { signal });\n  if (!res.ok) throw new Error(`HTTP ${res.status}`);\n  return await res.json();\n}",
         rubric: {
             precision: { criterion: '是否避免否定代码已经具备的健壮性措施', weight: 0.5 },
             reasoning: { criterion: '是否把确定事实与依赖业务上下文的增强建议分开', weight: 0.5 },
@@ -162,8 +162,8 @@ const INIT_CURATED_SAMPLES = [
     },
     {
         sample_id: 's011',
-        prompt: '审查以下代码',
-        context: 'function shippingFee(weight) {\n  if (weight > 30) return 999;\n  return weight * 7.35 + 12;\n}',
+        input: { inputKind: 'text', text: "审查以下代码\n\n```\nfunction shippingFee(weight) {\n  if (weight > 30) return 999;\n  return weight * 7.35 + 12;\n}\n```" },
+        reference: 'function shippingFee(weight) {\n  if (weight > 30) return 999;\n  return weight * 7.35 + 12;\n}',
         rubric: {
             maintainability: { criterion: '是否解释 30、999、7.35、12 的语义和变更风险', weight: 0.5 },
             actionability: { criterion: '是否给出命名、单位和规则归位的具体方案', weight: 0.5 },
@@ -175,8 +175,8 @@ const INIT_CURATED_SAMPLES = [
     },
     {
         sample_id: 's012',
-        prompt: '审查以下代码',
-        context: "function createUser(input) {\n  if (!input.email.includes('@')) throw new Error('bad email');\n  return db.users.insert(input);\n}\nfunction updateUser(input) {\n  if (!input.email.includes('@')) throw new Error('bad email');\n  return db.users.update(input);\n}",
+        input: { inputKind: 'text', text: "审查以下代码\n\n```\nfunction createUser(input) {\n  if (!input.email.includes('@')) throw new Error('bad email');\n  return db.users.insert(input);\n}\nfunction updateUser(input) {\n  if (!input.email.includes('@')) throw new Error('bad email');\n  return db.users.update(input);\n}\n```" },
+        reference: "function createUser(input) {\n  if (!input.email.includes('@')) throw new Error('bad email');\n  return db.users.insert(input);\n}\nfunction updateUser(input) {\n  if (!input.email.includes('@')) throw new Error('bad email');\n  return db.users.update(input);\n}",
         rubric: {
             maintainability: { criterion: '是否识别重复逻辑与未来规则不一致的风险', weight: 0.5 },
             actionability: { criterion: '是否提出职责清晰、易测试且不过度抽象的重构', weight: 0.5 },
@@ -188,8 +188,8 @@ const INIT_CURATED_SAMPLES = [
     },
     {
         sample_id: 's013',
-        prompt: '审查以下代码',
-        context: "export async function completeOrder(order, user) {\n  if (!user.admin && user.id !== order.userId) throw new Error('forbidden');\n  order.status = 'complete';\n  await db.orders.save(order);\n  await mail.send(user.email, renderReceipt(order));\n  metrics.increment('orders.complete');\n  return JSON.stringify(order);\n}",
+        input: { inputKind: 'text', text: "审查以下代码\n\n```\nexport async function completeOrder(order, user) {\n  if (!user.admin && user.id !== order.userId) throw new Error('forbidden');\n  order.status = 'complete';\n  await db.orders.save(order);\n  await mail.send(user.email, renderReceipt(order));\n  metrics.increment('orders.complete');\n  return JSON.stringify(order);\n}\n```" },
+        reference: "export async function completeOrder(order, user) {\n  if (!user.admin && user.id !== order.userId) throw new Error('forbidden');\n  order.status = 'complete';\n  await db.orders.save(order);\n  await mail.send(user.email, renderReceipt(order));\n  metrics.increment('orders.complete');\n  return JSON.stringify(order);\n}",
         rubric: {
             maintainability: { criterion: '是否识别职责耦合以及失败时产生的部分完成状态', weight: 0.5 },
             actionability: { criterion: '是否在拆分职责的同时保留事务和副作用顺序', weight: 0.5 },
@@ -201,8 +201,8 @@ const INIT_CURATED_SAMPLES = [
     },
     {
         sample_id: 's014',
-        prompt: '审查以下代码',
-        context: 'function renderReport(data, compact, includeHeader, sortDescending, useUtc) {\n  // formatting logic\n}',
+        input: { inputKind: 'text', text: "审查以下代码\n\n```\nfunction renderReport(data, compact, includeHeader, sortDescending, useUtc) {\n  // formatting logic\n}\n```" },
+        reference: 'function renderReport(data, compact, includeHeader, sortDescending, useUtc) {\n  // formatting logic\n}',
         rubric: {
             maintainability: { criterion: '是否解释调用点可读性和新增选项时的演进问题', weight: 0.5 },
             actionability: { criterion: '是否给出类型明确且可兼容默认值的参数设计', weight: 0.5 },
@@ -217,8 +217,8 @@ const INIT_CURATED_SAMPLES = [
     },
     {
         sample_id: 's015',
-        prompt: '审查以下代码',
-        context: 'export function clamp(value, min, max) {\n  return Math.min(max, Math.max(min, value));\n}',
+        input: { inputKind: 'text', text: "审查以下代码\n\n```\nexport function clamp(value, min, max) {\n  return Math.min(max, Math.max(min, value));\n}\n```" },
+        reference: 'export function clamp(value, min, max) {\n  return Math.min(max, Math.max(min, value));\n}',
         rubric: {
             precision: { criterion: '是否避免为了展示审查深度而虚构维护性问题', weight: 0.5 },
             proportionality: { criterion: '建议的复杂度是否与这个小型纯函数相称', weight: 0.5 },
@@ -230,8 +230,8 @@ const INIT_CURATED_SAMPLES = [
     },
     {
         sample_id: 's016',
-        prompt: '审查以下代码',
-        context: 'export async function listOrders(users) {\n  const rows = [];\n  for (const user of users) {\n    rows.push(...await db.orders.findByUser(user.id));\n  }\n  return rows;\n}',
+        input: { inputKind: 'text', text: "审查以下代码\n\n```\nexport async function listOrders(users) {\n  const rows = [];\n  for (const user of users) {\n    rows.push(...await db.orders.findByUser(user.id));\n  }\n  return rows;\n}\n```" },
+        reference: 'export async function listOrders(users) {\n  const rows = [];\n  for (const user of users) {\n    rows.push(...await db.orders.findByUser(user.id));\n  }\n  return rows;\n}',
         rubric: {
             performance: { criterion: '是否识别查询次数和串行延迟随用户数增长的问题', weight: 0.5 },
             actionability: { criterion: '是否给出符合数据库边界的批量读取方案', weight: 0.5 },
@@ -246,8 +246,8 @@ const INIT_CURATED_SAMPLES = [
     },
     {
         sample_id: 's017',
-        prompt: '审查以下代码',
-        context: 'export async function hydrate(ids) {\n  const result = [];\n  for (const id of ids) {\n    result.push(await fetch(`/items/${id}`).then(r => r.json()));\n  }\n  return result;\n}',
+        input: { inputKind: 'text', text: "审查以下代码\n\n```\nexport async function hydrate(ids) {\n  const result = [];\n  for (const id of ids) {\n    result.push(await fetch(`/items/${id}`).then(r => r.json()));\n  }\n  return result;\n}\n```" },
+        reference: 'export async function hydrate(ids) {\n  const result = [];\n  for (const id of ids) {\n    result.push(await fetch(`/items/${id}`).then(r => r.json()));\n  }\n  return result;\n}',
         rubric: {
             performance: { criterion: '是否同时看见串行瓶颈和无界并发的反向风险', weight: 0.5 },
             actionability: { criterion: '是否给出可调并发度、错误策略和顺序语义', weight: 0.5 },
@@ -262,8 +262,8 @@ const INIT_CURATED_SAMPLES = [
     },
     {
         sample_id: 's018',
-        prompt: '审查以下代码',
-        context: 'function commonIds(left, right) {\n  return left.filter(item => right.some(other => other.id === item.id));\n}',
+        input: { inputKind: 'text', text: "审查以下代码\n\n```\nfunction commonIds(left, right) {\n  return left.filter(item => right.some(other => other.id === item.id));\n}\n```" },
+        reference: 'function commonIds(left, right) {\n  return left.filter(item => right.some(other => other.id === item.id));\n}',
         rubric: {
             performance: { criterion: '是否准确分析时间复杂度而不是泛泛声称性能差', weight: 0.5 },
             actionability: { criterion: '是否根据唯一性和内存取舍选择合适索引结构', weight: 0.5 },
@@ -278,8 +278,8 @@ const INIT_CURATED_SAMPLES = [
     },
     {
         sample_id: 's019',
-        prompt: '审查以下代码',
-        context: 'const cache = new Map();\nexport async function resolveTenant(id) {\n  if (!cache.has(id)) cache.set(id, await loadTenant(id));\n  return cache.get(id);\n}',
+        input: { inputKind: 'text', text: "审查以下代码\n\n```\nconst cache = new Map();\nexport async function resolveTenant(id) {\n  if (!cache.has(id)) cache.set(id, await loadTenant(id));\n  return cache.get(id);\n}\n```" },
+        reference: 'const cache = new Map();\nexport async function resolveTenant(id) {\n  if (!cache.has(id)) cache.set(id, await loadTenant(id));\n  return cache.get(id);\n}',
         rubric: {
             performance: { criterion: '是否覆盖内存增长与并发 cache miss 两个独立问题', weight: 0.5 },
             actionability: { criterion: '是否给出与数据新鲜度和容量约束匹配的缓存策略', weight: 0.5 },
@@ -291,8 +291,8 @@ const INIT_CURATED_SAMPLES = [
     },
     {
         sample_id: 's020',
-        prompt: '审查以下代码',
-        context: 'export function indexById(items) {\n  const index = new Map();\n  for (const item of items) index.set(item.id, item);\n  return index;\n}',
+        input: { inputKind: 'text', text: "审查以下代码\n\n```\nexport function indexById(items) {\n  const index = new Map();\n  for (const item of items) index.set(item.id, item);\n  return index;\n}\n```" },
+        reference: 'export function indexById(items) {\n  const index = new Map();\n  for (const item of items) index.set(item.id, item);\n  return index;\n}',
         rubric: {
             precision: { criterion: '是否避免把正常的 O(n) 工作误报为性能缺陷', weight: 0.5 },
             reasoning: { criterion: '是否把重复键语义作为条件性业务问题而非确定 bug', weight: 0.5 },
@@ -304,5 +304,5 @@ const INIT_CURATED_SAMPLES = [
     },
 ];
 export function serializeInitSamples(count) {
-    return `${JSON.stringify(createEvalSampleSetDocument(INIT_CURATED_SAMPLES.slice(0, count)), null, 2)}\n`;
+    return `${JSON.stringify(createWorkflowSampleSetDocument(INIT_CURATED_SAMPLES.slice(0, count)), null, 2)}\n`;
 }

package/dist/dsh-plugin/index.js

@@ -99,12 +99,12 @@ async function studioUrl(invocation, state) {
         return state.serverUrl;
     const cwd = invocation.agent.session.header.cwd ?? process.cwd();
     const layout = projectLayout(cwd);
-    const { createReportServer } = await import('../studio/http/report-server.js');
+    const { createNextStudioServer } = await import('../studio/http/next-server.js');
     const { createNodeCoreContentStore, createNodeCoreRunArtifactStore, } = await import('../eval-workflows/artifact-store/index.js');
     const { createCoreStudioCatalog } = await import('../studio/application/core-run-catalog.js');
     const reportsDir = layout.evalDir;
     const contentStore = createNodeCoreContentStore(join(reportsDir, 'content'));
-    state.server = createReportServer({
+    state.server = createNextStudioServer({
         port: 0,
         coreStudioCatalog: createCoreStudioCatalog(createNodeCoreRunArtifactStore(reportsDir, {
             contentResolver: contentStore,
@@ -114,7 +114,7 @@ async function studioUrl(invocation, state) {
         observationsDir: layout.observeInboxDir,
         managedDir: layout.managedDir,
         conversationCatalog: state.catalog,
-        // DSH 宿主无收件箱页面入口（输出落盘路径），裁剪收件箱路由（#839 批次 0）。
+        // DSH 宿主无收件箱页面入口（输出落盘路径），裁剪收件箱页面与 API（#839 批次 0）。
         observationInbox: false,
     });
     state.serverUrl = await state.server.start();

package/dist/eval-runtime/adapters/command-executor.d.ts

@@ -3,7 +3,7 @@ import { type JsonValue } from '../../eval-core/contracts/index.js';
 import type { Executor, ExecutorCapabilities } from '../evaluation/contracts.js';
 /**
  * Façade-level subprocess exchange. Deliberately distinct from the sealed host-seam protocol
- * `omk.custom-command-exchange/v1`: a canonical invocation carries no runId, trialId, attemptId,
+ * `omk.custom-executor-exchange/v1`: a canonical invocation carries no runId, trialId, attemptId,
  * isolation key or execution plan digest, and this adapter never fabricates them. Hosts that need
  * plan-bound isolation, workspace overlays, native MCP config or mock interception must compose the
  * advanced execution seam instead.

package/dist/eval-runtime/adapters/command-executor.js

@@ -3,7 +3,7 @@ import { z } from 'zod';
 import { IdentifierSchema, JsonValueSchema, UsageRecordSchema, canonicalizeJson, digestCanonicalJson, } from '../../eval-core/contracts/index.js';
 /**
  * Façade-level subprocess exchange. Deliberately distinct from the sealed host-seam protocol
- * `omk.custom-command-exchange/v1`: a canonical invocation carries no runId, trialId, attemptId,
+ * `omk.custom-executor-exchange/v1`: a canonical invocation carries no runId, trialId, attemptId,
  * isolation key or execution plan digest, and this adapter never fabricates them. Hosts that need
  * plan-bound isolation, workspace overlays, native MCP config or mock interception must compose the
  * advanced execution seam instead.

package/dist/eval-runtime/evaluation/prepare.js

@@ -10,7 +10,7 @@ import { captureDataset, captureVariant, } from './capture-input.js';
 import { createEvaluationRuntime, EvaluationRuntimeAssemblyError, } from '../runtime.js';
 import { createEvaluationEngine as createCoreEvaluationEngine, } from '../../eval-core/engine/index.js';
 import { randomUUID, } from 'node:crypto';
-import { runPreparedEvaluation, EvaluationEventConsumptionError as AdvancedEvaluationEventConsumptionError, } from '../runner.js';
+import { runPreparedEvaluation, HostEvaluationEventConsumptionError as AdvancedEvaluationEventConsumptionError, } from '../runner.js';
 import { attachDefinition, corePreparedEvaluations, } from './result-state.js';
 import { captureEvaluators, } from './capture-evaluators.js';
 import { createGeneralDefinition, } from './definition.js';

package/dist/eval-runtime/evaluators/exact-match.d.ts

@@ -4,12 +4,14 @@ export declare const EXACT_MATCH_EVALUATOR_IMPLEMENTATION_ID: "omk.eval-runtime.
 export interface CreateExactMatchEvaluatorInput {
     readonly metricId?: string;
     readonly actualBindingId?: string;
+    readonly actualSourceKind?: 'output' | 'trace';
     readonly expectedBindingId?: string;
     readonly sessionIsolationKey?: string;
 }
 export declare function createExactMatchEvaluatorIdentity(input?: Readonly<{
     metricId?: string;
     actualBindingId?: string;
+    actualSourceKind?: 'output' | 'trace';
     expectedBindingId?: string;
 }>): RuntimeIdentity;
 /** Creates the deterministic exact-match Evaluator used by the convenience Definition builder. */

package/dist/eval-runtime/evaluators/exact-match.js

@@ -13,7 +13,7 @@ export function createExactMatchEvaluatorIdentity(input = {}) {
         implementationId: EXACT_MATCH_EVALUATOR_IMPLEMENTATION_ID,
         version: '1.0.0',
         capabilities: {
-            inputSourceKinds: ['expected', 'output'],
+            inputSourceKinds: ['expected', input.actualSourceKind ?? 'output'],
             metricValueTypes: ['boolean'],
             schemas: [],
         },
@@ -22,6 +22,7 @@ export function createExactMatchEvaluatorIdentity(input = {}) {
             metricId,
             actualBindingId,
             expectedBindingId,
+            ...(input.actualSourceKind === 'trace' ? { actualSourceKind: 'trace' } : {}),
         },
     });
 }
@@ -31,7 +32,7 @@ export function createExactMatchEvaluator(input = {}) {
     const actualBindingId = input.actualBindingId ?? 'actual';
     const expectedBindingId = input.expectedBindingId ?? 'expected';
     return createSameProcessEvaluatorAdapter({
-        identity: createExactMatchEvaluatorIdentity({ metricId, actualBindingId, expectedBindingId }),
+        identity: createExactMatchEvaluatorIdentity({ metricId, actualBindingId, expectedBindingId, actualSourceKind: input.actualSourceKind }),
         sessionIsolationKey: input.sessionIsolationKey ?? 'omk.eval-runtime.exact-match/v1',
         resourceLeases: { forRun: () => undefined },
         implementation: {

package/dist/eval-runtime/index.d.ts

@@ -24,3 +24,39 @@ export type { McpConfigAccess, McpConfigDescriptor, McpConfigInput, McpConfigLea
 export type { AllowedToolsInput, AllowedToolsPlan, } from './tool-policy.js';
 export type { WorkspaceAccess, WorkspaceDescriptor, WorkspaceInput, WorkspaceLease, WorkspaceOpenRequest, WorkspacePlan, WorkspaceProvider, } from './workspace.js';
 export type { CustomEvaluator, CustomEvaluatorBinding, CustomEvaluatorContent, CustomEvaluatorInvocation, CustomEvaluatorResult, Metric, } from './custom-evaluator.js';
+export { createExactMatchDefinition } from './builders/exact-match.js';
+export type { ExactMatchDefinitionBuilderInput, ExactMatchTarget, } from './builders/exact-match.js';
+export { createMeasurementPolicy } from './builders/policy.js';
+export type { MeasurementEventDeliveryInput, MeasurementFailurePolicyInput, MeasurementAttemptBudgetScopeInput, MeasurementBudgetPolicyInput, MeasurementBudgetScopeInput, MeasurementCachePolicyInput, MeasurementEvidencePolicyInput, MeasurementPolicyBuilderInput, MeasurementProviderCostLimitInput, MeasurementRetryBackoffInput, MeasurementRetryPolicyInput, MeasurementRunBudgetScopeInput, MeasurementStagePolicyInput, } from './builders/policy.js';
+export { createPairedComparisonDefinition } from './builders/paired-comparison.js';
+export type { EvaluationRuntimeTarget, PairedComparisonDefinitionBuilderInput, } from './builders/paired-comparison.js';
+export { EXACT_MATCH_EVALUATOR_IMPLEMENTATION_ID, createExactMatchEvaluator, createExactMatchEvaluatorIdentity, } from './evaluators/exact-match.js';
+export type { CreateExactMatchEvaluatorInput } from './evaluators/exact-match.js';
+export { createRubricJudgeCriterion, createRubricJudgeEvaluator, createRubricJudgeEvaluatorDefinition, createRubricJudgeEvaluatorIdentity, createRubricJudgeEvaluatorRegistration, createRubricJudgeInstrument, createRubricJudgeMetricDefinition, createRubricJudgeRuntimeConfig, rubricJudgeInstrumentId, } from './judges/rubric-judge.js';
+export type { CreateRubricJudgeEvaluatorInput, RubricJudgeEvaluatorBinding, RubricJudgeEvaluatorDefinitionBuilderInput, } from './judges/rubric-judge.js';
+export { createRubricJudgeEvaluationContext, createRubricJudgeKit, createRubricJudgeRegistration, } from './judges/rubric-kit.js';
+export type { CreateRubricJudgeKitInput, RubricJudgeKit, } from './judges/rubric-kit.js';
+export { HostEvaluationEventConsumptionError, runEvaluation, } from './runner.js';
+export type { EvaluationEventObserver, RunEvaluationInput, } from './runner.js';
+export { createInvokeExecutorIdentity, createRuntimeIdentity, createSessionExecutorIdentity, } from './identity.js';
+export type { InvokeExecutorIdentityDeclaration, RuntimeIdentityDeclaration, SessionExecutorIdentityDeclaration, } from './identity.js';
+export { EvaluationRuntimeAssemblyError, createEvaluationRuntime, } from './runtime.js';
+export type { CreateEvaluationRuntimeInput } from './runtime.js';
+export { createJsonExecutorAdapter, createJsonSessionExecutorAdapter, } from './adapters/json-executor.js';
+export type { CreateJsonExecutorAdapterInput, CreateJsonSessionExecutorAdapterInput, JsonExecutorInvocation, JsonExecutorInvocationResult, JsonExecutorSession, JsonSessionExecutorAttempt, JsonSessionExecutorContext, RuntimeValueParser, } from './adapters/json-executor.js';
+export { DEFAULT_SUBPROCESS_COMMAND_MAX_OUTPUT_BYTES, SUBPROCESS_COMMAND_EXCHANGE_SCHEMA_VERSION, createSubprocessCommandExecutor, } from './adapters/command-executor.js';
+export type { CreateSubprocessCommandExecutorInput, SubprocessCommandConfiguration, SubprocessCommandValueParser, } from './adapters/command-executor.js';
+export type { OmkLlmJudgeEffort, OmkLlmJudgeInvocationPort, OmkLlmJudgeInvocationRequest, OmkLlmJudgeInvocationResult, } from './judges/invocation.js';
+export { RuntimeConformanceError, assertExecutorConformance, runExecutorConformance, } from './conformance/executor.js';
+export type { ExecutorConformanceProbeInput, ExecutorConformanceResult, } from './conformance/executor.js';
+export { createNodeEvaluationClock } from './clock.js';
+export { INVOKE_JSON_INPUT_SCHEMA, INVOKE_JSON_OUTPUT_SCHEMA, INVOKE_JSON_TRACE_SCHEMA, SESSION_JSON_INPUT_SCHEMA, SESSION_JSON_OUTPUT_SCHEMA, SESSION_JSON_TRACE_SCHEMA, } from './identity.js';
+export type { EvaluationRuntimeSupportPorts, RuntimePortRegistration, } from './runtime.js';
+export { createExecutorFnAdapter } from './adapters/executor-fn.js';
+export type { CreateExecutorFnAdapterInput, ExecResult, ExecutorFn, ExecutorFnInputMapper, ExecutorFnResultMapper, ExecutorInput, } from './adapters/executor-fn.js';
+export { createSameProcessEvaluatorAdapter, createSameProcessExecutorAdapter, } from './adapters/same-process.js';
+export type { CreateSameProcessEvaluatorAdapterInput, CreateSameProcessExecutorAdapterInput, SameProcessEvaluatorImplementation, SameProcessExecutorImplementation, SameProcessOperationScope, SameProcessResourceLeaseAccess, SameProcessRunScope, } from './adapters/same-process.js';
+export { RUBRIC_JUDGE_BINDINGS, RUBRIC_JUDGE_CONTEXT_SCHEMA, RUBRIC_JUDGE_CONTEXT_SCHEMA_VERSION, RUBRIC_JUDGE_EVALUATOR_IMPLEMENTATION_ID, RUBRIC_JUDGE_EVIDENCE_SCHEMA, RUBRIC_JUDGE_EVIDENCE_SCHEMA_VERSION, RUBRIC_JUDGE_INSTRUMENT_SCHEMA, RUBRIC_JUDGE_INSTRUMENT_SCHEMA_VERSION, } from './judges/rubric-contracts.js';
+export type { RubricJudgeConfig, RubricJudgeCriterion, RubricJudgeInstrument, RubricJudgeRuntimeConfig, RubricJudgeTracePolicy, } from './judges/rubric-contracts.js';
+export { SOURCE_NEUTRAL_TRACE_SCHEMA_DESCRIPTOR, SOURCE_NEUTRAL_TRACE_SCHEMA_VERSION, SOURCE_NEUTRAL_TRACE_WITHOUT_MOCKS_SCHEMA_DESCRIPTOR, SourceNeutralMockStatsSchema, SourceNeutralTraceSchema, SourceNeutralTraceWithoutMocksSchema, attachSourceNeutralMockStats, parseSourceNeutralTrace, } from './traces/source-neutral.js';
+export type { SourceNeutralMockStats, SourceNeutralTrace, } from './traces/source-neutral.js';

package/dist/eval-runtime/index.js

@@ -4,3 +4,24 @@ export { EVALUATION_RESULT_MEDIA_TYPE, EvaluationResultStoreError, loadEvaluatio
 export { EXECUTED_EVALUATION_MEDIA_TYPE, ExecutedEvaluationStoreError, loadExecutedEvaluation, saveExecutedEvaluation, } from './executed-store.js';
 export { RUNTIME_CHECK_RESULT_SCHEMA_VERSION, checkRuntime, } from './conformance/runtime.js';
 export { MOCK_INTERCEPTION_PLAN_MEDIA_TYPE } from './mock-interception.js';
+// Imperative builders for assembling Definitions, Policies, and Evaluators directly.
+export { createExactMatchDefinition } from './builders/exact-match.js';
+export { createMeasurementPolicy } from './builders/policy.js';
+export { createPairedComparisonDefinition } from './builders/paired-comparison.js';
+export { EXACT_MATCH_EVALUATOR_IMPLEMENTATION_ID, createExactMatchEvaluator, createExactMatchEvaluatorIdentity, } from './evaluators/exact-match.js';
+export { createRubricJudgeCriterion, createRubricJudgeEvaluator, createRubricJudgeEvaluatorDefinition, createRubricJudgeEvaluatorIdentity, createRubricJudgeEvaluatorRegistration, createRubricJudgeInstrument, createRubricJudgeMetricDefinition, createRubricJudgeRuntimeConfig, rubricJudgeInstrumentId, } from './judges/rubric-judge.js';
+export { createRubricJudgeEvaluationContext, createRubricJudgeKit, createRubricJudgeRegistration, } from './judges/rubric-kit.js';
+// Host assembly SPI (formerly eval-runtime/advanced).
+export { HostEvaluationEventConsumptionError, runEvaluation, } from './runner.js';
+export { createInvokeExecutorIdentity, createRuntimeIdentity, createSessionExecutorIdentity, } from './identity.js';
+export { EvaluationRuntimeAssemblyError, createEvaluationRuntime, } from './runtime.js';
+export { createJsonExecutorAdapter, createJsonSessionExecutorAdapter, } from './adapters/json-executor.js';
+export { DEFAULT_SUBPROCESS_COMMAND_MAX_OUTPUT_BYTES, SUBPROCESS_COMMAND_EXCHANGE_SCHEMA_VERSION, createSubprocessCommandExecutor, } from './adapters/command-executor.js';
+export { RuntimeConformanceError, assertExecutorConformance, runExecutorConformance, } from './conformance/executor.js';
+export { createNodeEvaluationClock } from './clock.js';
+export { INVOKE_JSON_INPUT_SCHEMA, INVOKE_JSON_OUTPUT_SCHEMA, INVOKE_JSON_TRACE_SCHEMA, SESSION_JSON_INPUT_SCHEMA, SESSION_JSON_OUTPUT_SCHEMA, SESSION_JSON_TRACE_SCHEMA, } from './identity.js';
+export { createExecutorFnAdapter } from './adapters/executor-fn.js';
+export { createSameProcessEvaluatorAdapter, createSameProcessExecutorAdapter, } from './adapters/same-process.js';
+// Wire contracts (formerly eval-runtime/contracts).
+export { RUBRIC_JUDGE_BINDINGS, RUBRIC_JUDGE_CONTEXT_SCHEMA, RUBRIC_JUDGE_CONTEXT_SCHEMA_VERSION, RUBRIC_JUDGE_EVALUATOR_IMPLEMENTATION_ID, RUBRIC_JUDGE_EVIDENCE_SCHEMA, RUBRIC_JUDGE_EVIDENCE_SCHEMA_VERSION, RUBRIC_JUDGE_INSTRUMENT_SCHEMA, RUBRIC_JUDGE_INSTRUMENT_SCHEMA_VERSION, } from './judges/rubric-contracts.js';
+export { SOURCE_NEUTRAL_TRACE_SCHEMA_DESCRIPTOR, SOURCE_NEUTRAL_TRACE_SCHEMA_VERSION, SOURCE_NEUTRAL_TRACE_WITHOUT_MOCKS_SCHEMA_DESCRIPTOR, SourceNeutralMockStatsSchema, SourceNeutralTraceSchema, SourceNeutralTraceWithoutMocksSchema, attachSourceNeutralMockStats, parseSourceNeutralTrace, } from './traces/source-neutral.js';

package/dist/eval-runtime/runner.d.ts

@@ -30,12 +30,12 @@ export interface RunPreparedEvaluationInput {
     readonly eventBufferCapacity?: number;
     readonly onEvent?: EvaluationEventObserver;
 }
-export declare class EvaluationEventConsumptionError extends Error {
+export declare class HostEvaluationEventConsumptionError extends Error {
     readonly code: 'EVAL_RUNTIME_EVENT_OBSERVER_FAILED' | 'EVAL_RUNTIME_EVENT_STREAM_FAILED';
     readonly runResult?: EvaluationRunResult;
     readonly cause: unknown;
     constructor(input: Readonly<{
-        code: EvaluationEventConsumptionError['code'];
+        code: HostEvaluationEventConsumptionError['code'];
         message: string;
         cause: unknown;
         runResult?: EvaluationRunResult;

package/dist/eval-runtime/runner.js

@@ -1,11 +1,11 @@
 import { createEvaluationEngine as createCoreEvaluationEngine } from '../eval-core/engine/index.js';
-export class EvaluationEventConsumptionError extends Error {
+export class HostEvaluationEventConsumptionError extends Error {
     code;
     runResult;
     cause;
     constructor(input) {
         super(input.message);
-        this.name = 'EvaluationEventConsumptionError';
+        this.name = 'HostEvaluationEventConsumptionError';
         this.code = input.code;
         this.cause = input.cause;
         this.runResult = input.runResult;
@@ -62,7 +62,7 @@ async function consumeEvaluationRun(start, input) {
             throw resultOutcome.reason;
         const result = resultOutcome.value;
         if (observerFailed) {
-            throw new EvaluationEventConsumptionError({
+            throw new HostEvaluationEventConsumptionError({
                 code: 'EVAL_RUNTIME_EVENT_OBSERVER_FAILED',
                 message: 'Evaluation event observer 执行失败；评测保持 Core 终态并完成清理。',
                 cause: observerFailure,
@@ -70,7 +70,7 @@ async function consumeEvaluationRun(start, input) {
             });
         }
         if (streamFailed) {
-            throw new EvaluationEventConsumptionError({
+            throw new HostEvaluationEventConsumptionError({
                 code: 'EVAL_RUNTIME_EVENT_STREAM_FAILED',
                 message: 'Evaluation event stream 消费失败；评测已取消并完成清理。',
                 cause: streamFailure,

package/dist/eval-workflows/hosts/adapters/anthropic/api.js

@@ -1,3 +1,5 @@
+import { projectAnthropicApiInput } from './input.js';
+import { STATELESS_API_SAMPLE_INPUT_POLICY } from '../shared/sample-input.js';
 import { z } from 'zod';
 import { RuntimeIdentitySchema, canonicalizeJson, deepFreezeCanonicalJson, digestCanonicalJson, } from '../../../../eval-core/contracts/index.js';
 import { ExecutionPortFailure, } from '../../../../eval-core/execution/index.js';
@@ -81,7 +83,8 @@ function identityManifest(configuration, target, policy) {
             facetId: 'adapter.input-projection',
             value: {
                 directoryEntrypoint: 'SKILL.md',
-                promptTransport: 'messages-user-text',
+                authoredInput: STATELESS_API_SAMPLE_INPUT_POLICY,
+                promptTransport: 'messages-text-or-native-history',
                 supportingFiles: 'canonical-user-envelope',
                 systemInstructions: 'top-level-system',
                 version: RESOURCE_PROFILE.promptSchemaVersion,
@@ -153,11 +156,8 @@ function requestBody(target, policy, runState, trialState) {
     return JSON.stringify({
         model: target.binding.qualification.model,
         max_tokens: policy.maxOutputTokens,
-        messages: [{ role: 'user', content: trialState.prompt }],
+        ...projectAnthropicApiInput(runState, trialState),
         stream: false,
-        ...(runState.systemInstructions === undefined
-            ? {}
-            : { system: runState.systemInstructions }),
         ...(target.binding.qualification.effort === undefined
             ? {}
             : { output_config: { effort: target.binding.qualification.effort } }),