Validate run input types

Author: kael-odin

README.md

@@ -14,7 +14,7 @@ Chinese documentation: [README_CN.md](./README_CN.md).
   - `Result/PushData`
   - `Log/Debug`, `Log/Info`, `Log/Warn`, `Log/Error`
 - Runtime input injection from `input_schema.json` defaults, `--input`, or `--json`
-- Run input validation for required `input_schema.json` fields before the worker starts
+- Run input validation for required fields and declared value types in `input_schema.json` before the worker starts
 - Platform environment variables:
   - `ChromeWs`
   - `CDP_ENDPOINT` / `BROWSER_WS_ENDPOINT`
@@ -93,7 +93,7 @@ Validation checks:
 
 CoreClaw installs dependencies from `requirements.txt`, `package.json`, or `go.mod` after upload. The CLI therefore rejects workers that rely on locally installed SDK packages but do not declare those packages for the cloud installer.
 
-At run time, the CLI also validates the actual input assembled from defaults, `--input`, or `--json`. If a field marked `"required": true` is missing or empty, the command fails before creating run artifacts or starting the worker, matching CoreClaw's form-level launch behavior.
+At run time, the CLI also validates the actual input assembled from defaults, `--input`, or `--json`. If a field marked `"required": true` is missing or empty, or if a declared input field has the wrong JSON type, the command fails before creating run artifacts or starting the worker, matching CoreClaw's form-level launch behavior.
 
 CoreClaw's docs describe `output_schema.json` for upload-ready projects, but the current platform still accepts older workers without it. The CLI treats a missing `output_schema.json` as a warning, not a blocker. Local `export.ndjson` keeps the full raw result rows when no output schema exists.
 
@@ -128,7 +128,7 @@ The run starts a local CoreClaw SDK gRPC server on `127.0.0.1:20086`, then execu
 
 Use `--timeout-ms` to cap the whole worker process and `--idle-timeout-ms` to stop a worker that has stopped producing output but still has open Node/Python/Go handles. Durations accept milliseconds, `s`, or `m`.
 
-If the input schema marks a field as required, local runs require a non-empty value for that field. Use `--input input.json` or `--json '{"field":"value"}'` when the schema does not provide a default.
+If the input schema marks a field as required, local runs require a non-empty value for that field. Declared fields must also match their schema type, for example `integer` must be an integer, `boolean` must be a boolean, and `array` must be a JSON array. Use `--input input.json` or `--json '{"field":"value"}'` when the schema does not provide a default.
 
 Use `--min-results` for real worker smoke tests. Some existing workers can exit with code `0` after logging an upstream or browser error, so result count is the reliable success gate.
 

README_CN.md

@@ -12,7 +12,7 @@ CoreClaw 官方开发者文档描述了上传就绪的 worker 项目结构、平
   - `Result/PushData`
   - `Log/Debug`、`Log/Info`、`Log/Warn`、`Log/Error`
 - 从 `input_schema.json` 默认值、`--input` 或 `--json` 注入运行输入。
-- worker 启动前校验实际输入是否满足 `input_schema.json` 的 required 字段。
+- worker 启动前校验实际输入是否满足 `input_schema.json` 的 required 字段和声明类型。
 - 平台环境变量：
   - `ChromeWs`
   - `CDP_ENDPOINT` / `BROWSER_WS_ENDPOINT`
@@ -92,7 +92,7 @@ node ./bin/coreclaw.js validate ./examples/node-hello
 
 CoreClaw 上传后会从 `requirements.txt`、`package.json` 或 `go.mod` 安装依赖。因此 CLI 会拒绝那些本地机器因为已安装 SDK 包而能运行、但云端安装文件没有声明这些包的 worker。
 
-运行时，CLI 还会校验由默认值、`--input` 或 `--json` 拼出的实际输入。如果某个字段标记了 `"required": true`，但本次输入缺失或为空，命令会在创建 run 产物和启动 worker 前失败，贴近 CoreClaw 表单层的启动行为。
+运行时，CLI 还会校验由默认值、`--input` 或 `--json` 拼出的实际输入。如果某个字段标记了 `"required": true` 但本次输入缺失或为空，或者声明字段的 JSON 类型不匹配，命令会在创建 run 产物和启动 worker 前失败，贴近 CoreClaw 表单层的启动行为。
 
 官方文档把 `output_schema.json` 描述为上传就绪项目文件，但当前平台仍兼容没有 `output_schema.json` 的老 worker。因此 CLI 把缺失 `output_schema.json` 作为 warning，而不是阻塞错误。没有 output schema 时，本地 `export.ndjson` 会保留完整原始结果行。
 
@@ -127,7 +127,7 @@ node ./bin/coreclaw.js run ./browser-worker --captcha-solver --require-captcha-s
 
 `--timeout-ms` 用于限制整个 worker 进程运行时间；`--idle-timeout-ms` 用于停止已经不再输出但仍有 Node/Python/Go handle 未退出的 worker。时长支持毫秒、`s` 和 `m`。
 
-如果 input schema 把某个字段标记为 required，本地 run 也要求该字段有非空值。schema 没有提供 default 时，使用 `--input input.json` 或 `--json '{"field":"value"}'` 传入。
+如果 input schema 把某个字段标记为 required，本地 run 也要求该字段有非空值。声明字段还必须匹配 schema 类型，例如 `integer` 必须是整数，`boolean` 必须是布尔值，`array` 必须是 JSON 数组。schema 没有提供 default 时，使用 `--input input.json` 或 `--json '{"field":"value"}'` 传入。
 
 真实 worker 冒烟测试应使用 `--min-results`。有些 worker 会在上游或浏览器失败后仍以 exit code `0` 退出，因此结果行数才是更可靠的成功门槛。
 

src/runtime/input.js

@@ -63,11 +63,23 @@ export function inputSchemaInputIssues(input, schema) {
 
   const issues = [];
   for (const property of schema.properties) {
-    if (!property || property.required !== true || typeof property.name !== 'string') {
+    if (!property || typeof property.name !== 'string') {
       continue;
     }
-    if (!Object.prototype.hasOwnProperty.call(input, property.name) || isEmptyInputValue(input[property.name])) {
+    if (!Object.prototype.hasOwnProperty.call(input, property.name)) {
+      if (property.required === true) {
+        issues.push(`required field "${property.name}" is missing or empty`);
+      }
+      continue;
+    }
+
+    if (property.required === true && isEmptyInputValue(input[property.name])) {
       issues.push(`required field "${property.name}" is missing or empty`);
+      continue;
+    }
+
+    if (!inputValueMatchesType(input[property.name], property.type)) {
+      issues.push(`field "${property.name}" must be ${inputTypeLabel(property.type)}`);
     }
   }
   return issues;
@@ -120,6 +132,41 @@ function isEmptyInputValue(value) {
   return false;
 }
 
+function inputValueMatchesType(value, type) {
+  switch (normalizeInputType(type)) {
+    case 'string':
+      return typeof value === 'string';
+    case 'integer':
+      return Number.isInteger(value);
+    case 'boolean':
+      return typeof value === 'boolean';
+    case 'array':
+      return Array.isArray(value);
+    case 'object':
+      return value !== null && typeof value === 'object' && !Array.isArray(value);
+    default:
+      return true;
+  }
+}
+
+function inputTypeLabel(type) {
+  const normalized = normalizeInputType(type);
+  if (normalized === 'integer') {
+    return 'an integer';
+  }
+  if (normalized === 'array' || normalized === 'object') {
+    return `an ${normalized}`;
+  }
+  return `a ${normalized}`;
+}
+
+function normalizeInputType(type) {
+  if (type === 'number') {
+    return 'integer';
+  }
+  return type;
+}
+
 function singularize(key) {
   if (key.endsWith('ies')) {
     return `${key.slice(0, -3)}y`;

test/run.test.js

@@ -185,6 +185,30 @@ main().catch((error) => {
   assert.equal(fs.existsSync(path.join(dir, '.coreclaw')), false);
 });
 
+test('runCommand fails before creating run artifacts when input type is invalid', async () => {
+  const dir = createNodeFixture(`
+const coresdk = require('./sdk')
+async function main() {
+  await coresdk.result.pushData({ ok: true })
+}
+main().catch((error) => {
+  console.error(error)
+  process.exit(1)
+})
+`);
+
+  await assert.rejects(
+    () => runCommand(dir, {
+      node: process.execPath,
+      json: '{"items":"not a list"}',
+      tmpHook: false,
+    }),
+    (error) => error instanceof CliError && /field "items" must be an array/.test(error.message),
+  );
+
+  assert.equal(fs.existsSync(path.join(dir, '.coreclaw')), false);
+});
+
 test('runCommand fails upload parity when proxy usage is required but worker does not use it', async () => {
   const dir = createNodeFixture(`
 const coresdk = require('./sdk')

test/schema.test.js

@@ -75,6 +75,44 @@ test('validates actual run input against required schema fields', () => {
   );
 });
 
+test('validates actual run input types against input schema fields', () => {
+  const schema = {
+    properties: [
+      { name: 'keyword', type: 'string' },
+      { name: 'limit', type: 'integer' },
+      { name: 'legacyLimit', type: 'number' },
+      { name: 'enabled', type: 'boolean' },
+      { name: 'items', type: 'array' },
+      { name: 'options', type: 'object' },
+    ],
+  };
+
+  assert.deepEqual(inputSchemaInputIssues({
+    keyword: 'coreclaw',
+    limit: 3,
+    legacyLimit: 4,
+    enabled: false,
+    items: [],
+    options: {},
+    extra: 'allowed',
+  }, schema), []);
+  assert.deepEqual(inputSchemaInputIssues({
+    keyword: 123,
+    limit: 1.5,
+    legacyLimit: '4',
+    enabled: 'false',
+    items: {},
+    options: [],
+  }, schema), [
+    'field "keyword" must be a string',
+    'field "limit" must be an integer',
+    'field "legacyLimit" must be an integer',
+    'field "enabled" must be a boolean',
+    'field "items" must be an array',
+    'field "options" must be an object',
+  ]);
+});
+
 test('validates output schema', () => {
   const issues = validateOutputSchema([
     { name: 'url', type: 'string', description: 'URL' },