@reconcrap/boss-recommend-mcp 1.3.7 → 1.3.8

package/README.md

@@ -201,10 +201,12 @@ node src/cli.js chat run --job "算法工程师" --start-from unread --criteria
 
 - CLI：
   - `boss-recommend-mcp chat health-check`
+  - `boss-recommend-mcp chat prepare-run`
   - `boss-recommend-mcp chat run --job "算法工程师" --start-from unread --targetCount 20 --criteria "有 AI Agent 经验"`（后台启动，不自动轮询）
   - `boss-recommend-mcp chat start-run|get-run|pause-run|resume-run|cancel-run`
 - MCP：
   - `boss_chat_health_check`
+  - `prepare_boss_chat_run`
   - `start_boss_chat_run`
   - `get_boss_chat_run`
   - `pause_boss_chat_run`
@@ -213,13 +215,14 @@ node src/cli.js chat run --job "算法工程师" --start-from unread --criteria
 
 chat-only 交互建议：
 
-- 先调用一次 `start_boss_chat_run`（可不带参数），服务会先导航到 `https://www.zhipin.com/web/chat/index` 并返回 `NEED_INPUT`，其中包含岗位 `job_options` 与待补字段。
-- 然后基于 `job_options` 让用户选择 `job`，并补齐 `start_from`、`target_count`、`criteria` 后再次调用 `start_boss_chat_run` 启动任务。
+- 先调用一次 `prepare_boss_chat_run`（可不带参数），服务会先导航到 `https://www.zhipin.com/web/chat/index` 并返回 `NEED_INPUT`，其中包含岗位 `job_options` 与待补字段。
+- 然后基于 `job_options` 让用户选择 `job`，并补齐 `start_from`、`target_count`、`criteria` 后调用 `start_boss_chat_run` 启动任务。
 - `target_count` 支持正整数；若用户给出 `全部候选人` / `所有候选人`，会自动按不限（扫到底）处理。
 
 Trae-CN / 长对话防循环建议：
 
-- 固定流程：`boss_chat_health_check` -> `start_boss_chat_run(空参可)` -> 一次性补齐 `job/start_from/target_count/criteria` -> 再次 `start_boss_chat_run`。
+- 固定流程：`boss_chat_health_check` -> `prepare_boss_chat_run(空参可)` -> 一次性补齐 `job/start_from/target_count/criteria` -> `start_boss_chat_run`。
+- `start_boss_chat_run` 的工具 schema 已把 `job/start_from/target_count/criteria` 标记为必填；不要用它获取岗位列表。
 - `start_boss_chat_run` 返回 `ACCEPTED` 后直接结束当前回合，不要自动轮询。
 - 缺参或校验失败时，一次性列出全部缺失/错误项，避免重复同一句提示触发宿主“陷入循环”保护。
 - 仅当用户明确要求“查进度”时再调用 `get_boss_chat_run`。

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@reconcrap/boss-recommend-mcp",
-  "version": "1.3.7",
+  "version": "1.3.8",
   "description": "Unified MCP pipeline for recommend-page filtering and screening on Boss Zhipin",
   "keywords": [
     "boss",

package/skills/boss-chat/README.md

@@ -13,7 +13,7 @@ Please run a Boss chat-only task (do not switch to recommend flow).
 
 Execution order:
 1) Call boss_chat_health_check.
-2) Call start_boss_chat_run once (empty params allowed) to fetch job_options and missing fields.
+2) Call prepare_boss_chat_run once (empty params allowed) to fetch job_options and missing fields.
 3) Ask for these required fields in one shot: job, start_from (unread/all), target_count, criteria.
 4) After user reply, call start_boss_chat_run exactly once to start the run.
 5) If ACCEPTED, reply only with run_id and "task started"; no auto polling.
@@ -21,6 +21,7 @@ Execution order:
 Anti-loop rules:
 - Do not repeat the same sentence across turns.
 - On validation errors, list all missing/invalid fields once.
+- Do not use start_boss_chat_run for preflight. It is only for the final start call and must include job/start_from/target_count/criteria.
 - Do not call start_boss_chat_run repeatedly in one turn.
 - Do not call get_boss_chat_run unless user explicitly asks for progress.
 
@@ -30,4 +31,5 @@ target_count mapping:
 - `全部候选人` / `所有候选人` must also be treated as unlimited.
 - Always write the argument key as `target_count`.
 - For unlimited mode, send `"target_count": "all"` in the tool call.
+- If start_boss_chat_run returns NEED_INPUT for `target_count`, the previous tool call omitted the argument. Retry once using `next_call_example` and include `"target_count": "all"` or a positive integer.
 ```

package/skills/boss-chat/SKILL.md

@@ -14,6 +14,7 @@ description: "Use when users want Boss chat-page screening/outreach via the bund
 ## Tool Routing
 
 - 健康检查：`boss_chat_health_check`
+- 预备并获取岗位列表：`prepare_boss_chat_run`
 - 启动异步任务：`start_boss_chat_run`
 - 查询进度：`get_boss_chat_run`
 - 暂停：`pause_boss_chat_run`
@@ -29,7 +30,6 @@ description: "Use when users want Boss chat-page screening/outreach via the bund
 
 可选：
 
-- `target_count`
 - `profile`（默认 `default`）
 - `port`
 - `dry_run`
@@ -57,7 +57,9 @@ description: "Use when users want Boss chat-page screening/outreach via the bund
 - 当用户选择“扫到底/全部候选人/所有候选人”时，调用参数统一写：`"target_count": "all"`。
 - 禁止 agent 自行补全 `job/start_from/criteria` 并直接执行，必须由用户明确给出或确认。
 - chat-only 启动流程必须先进入聊天页并拉取岗位列表，再让用户从列表中选择 `job`。
-- 允许先用空参调用 `start_boss_chat_run` 触发 `NEED_INPUT`；若返回了 `job_options`，必须完整展示所有岗位选项给用户确认。
+- 必须先用空参调用 `prepare_boss_chat_run` 获取 `job_options`；不要用 `start_boss_chat_run` 做预备调用。
+- `start_boss_chat_run` 只能用于真正启动，必须一次性传齐 `job` / `start_from` / `target_count` / `criteria`。
+- 若 `start_boss_chat_run` 返回 `NEED_INPUT` 且 `missing_fields` 包含 `target_count`，说明你没有把用户选择写入工具参数；下一次调用必须照 `next_call_example` 原样补上 `"target_count": "all"` 或正整数，不要重复空调用。
 - 默认不自动轮询；只有用户要求查进度时才调用 `get_boss_chat_run`。
 - `start_boss_chat_run` 返回 `ACCEPTED` 后，默认立即结束当前回合，不得主动连续调用 `get_boss_chat_run`。
 - 只有当用户明确给出“轮询频率/间隔”（例如“每30分钟查一次”）时，才允许按该频率查询进度。
@@ -73,6 +75,6 @@ description: "Use when users want Boss chat-page screening/outreach via the bund
 ## Response Style
 
 - 用结构化中文。
-- 首轮建议先调用一次 `start_boss_chat_run`（可空参）获取 `job_options` 与 `pending_questions`。
+- 首轮建议先调用一次 `prepare_boss_chat_run`（可空参）获取 `job_options` 与 `pending_questions`。
 - 缺参时必须逐项确认：`job`（来自岗位列表）、`start_from`（`unread|all`）、`target_count`、`criteria`。
 - 若健康检查失败，明确提示共享配置文件 `screening-config.json` 不可用。

package/src/boss-chat.js

@@ -508,7 +508,23 @@ export async function startBossChatRun({ workspaceRoot, input = {} }) {
 }
 
 export async function prepareBossChatRun({ workspaceRoot, input = {} }) {
-  return (await spawnBossChatCli({ workspaceRoot, command: "prepare-run", input })).payload;
+  const payload = (await spawnBossChatCli({ workspaceRoot, command: "prepare-run", input })).payload;
+  if (payload?.status !== "NEED_INPUT") return payload;
+
+  const missingFields = getMissingBossChatStartFields(input);
+  const pendingQuestions = Array.isArray(payload?.pending_questions)
+    ? payload.pending_questions.filter((item) => (
+      missingFields.length === 0 || missingFields.includes(String(item?.field || ""))
+    ))
+    : [];
+  const nextCallExample = buildNextCallExample(input, missingFields);
+  return {
+    ...payload,
+    required_fields: CHAT_REQUIRED_FIELDS.slice(),
+    missing_fields: missingFields,
+    pending_questions: normalizePendingQuestions(pendingQuestions),
+    ...(nextCallExample ? { next_call_example: nextCallExample } : {})
+  };
 }
 
 export async function getBossChatRun({ workspaceRoot, input = {} }) {

package/src/cli.js

@@ -21,6 +21,7 @@ import {
   getBossChatHealthCheck,
   getBossChatRun,
   pauseBossChatRun,
+  prepareBossChatRun,
   resumeBossChatRun,
   startBossChatRun
 } from "./boss-chat.js";
@@ -1360,6 +1361,14 @@ async function runBossChatCliCommand(subcommand, options = {}) {
     return;
   }
 
+  if (subcommand === "prepare-run") {
+    printJson(await prepareBossChatRun({
+      workspaceRoot,
+      input: buildBossChatCliInput(options)
+    }));
+    return;
+  }
+
   if (subcommand === "run") {
     printJson(await startBossChatRun({
       workspaceRoot,

package/src/index.js

@@ -13,6 +13,7 @@ import {
   getBossChatHealthCheck,
   getBossChatRun,
   pauseBossChatRun,
+  prepareBossChatRun,
   resumeBossChatRun,
   startBossChatRun
 } from "./boss-chat.js";
@@ -51,6 +52,7 @@ const TOOL_RUN_FEATURED_CALIBRATION = "run_featured_calibration";
 const TOOL_GET_FEATURED_CALIBRATION_STATUS = "get_featured_calibration_status";
 const TOOL_RUN_RECOMMEND_SELF_HEAL = "run_recommend_self_heal";
 const TOOL_BOSS_CHAT_HEALTH_CHECK = "boss_chat_health_check";
+const TOOL_BOSS_CHAT_PREPARE_RUN = "prepare_boss_chat_run";
 const TOOL_BOSS_CHAT_START_RUN = "start_boss_chat_run";
 const TOOL_BOSS_CHAT_GET_RUN = "get_boss_chat_run";
 const TOOL_BOSS_CHAT_PAUSE_RUN = "pause_boss_chat_run";
@@ -413,8 +415,8 @@ function createRunInputSchema() {
   };
 }
 
-function createBossChatStartInputSchema() {
-  return {
+function createBossChatStartInputSchema({ requireFullInput = false } = {}) {
+  const schema = {
     type: "object",
     properties: {
       profile: {
@@ -459,6 +461,10 @@ function createBossChatStartInputSchema() {
     },
     additionalProperties: false
   };
+  if (requireFullInput) {
+    schema.required = ["job", "start_from", "target_count", "criteria"];
+  }
+  return schema;
 }
 
 function createRunFeaturedCalibrationInputSchema() {
@@ -610,10 +616,15 @@ function createToolsSchema() {
       }
     },
     {
-      name: TOOL_BOSS_CHAT_START_RUN,
-      description: "异步启动一次 boss-chat 任务。若缺少必填参数会先返回 NEED_INPUT（含岗位列表与待确认字段）。",
+      name: TOOL_BOSS_CHAT_PREPARE_RUN,
+      description: "预备一次 boss-chat 任务：只导航聊天页并返回岗位列表与待补字段，不会启动任务。用它先获取 job_options。",
       inputSchema: createBossChatStartInputSchema()
     },
+    {
+      name: TOOL_BOSS_CHAT_START_RUN,
+      description: "异步启动一次 boss-chat 任务。必须一次性提供 job、start_from、target_count、criteria；扫到底请传 target_count=\"all\"。",
+      inputSchema: createBossChatStartInputSchema({ requireFullInput: true })
+    },
     {
       name: TOOL_BOSS_CHAT_GET_RUN,
       description: "查询 boss-chat run_id 的当前状态。",
@@ -1776,6 +1787,10 @@ function handleBossChatHealthCheckTool(workspaceRoot, args) {
   return getBossChatHealthCheck(workspaceRoot, args);
 }
 
+async function handleBossChatPrepareRunTool({ workspaceRoot, args }) {
+  return prepareBossChatRun({ workspaceRoot, input: args });
+}
+
 async function handleBossChatStartRunTool({ workspaceRoot, args }) {
   return startBossChatRun({ workspaceRoot, input: args });
 }
@@ -1859,7 +1874,7 @@ async function handleRequest(message, workspaceRoot) {
       }
     }
 
-    if (toolName === TOOL_BOSS_CHAT_START_RUN) {
+    if ([TOOL_BOSS_CHAT_PREPARE_RUN, TOOL_BOSS_CHAT_START_RUN].includes(toolName)) {
       const inputError = validateBossChatStartArgs(args);
       if (inputError) {
         return createJsonRpcError(id, -32602, inputError);
@@ -1901,6 +1916,8 @@ async function handleRequest(message, workspaceRoot) {
         payload = await handleRunRecommendSelfHealTool({ workspaceRoot, args });
       } else if (toolName === TOOL_BOSS_CHAT_HEALTH_CHECK) {
         payload = handleBossChatHealthCheckTool(workspaceRoot, args);
+      } else if (toolName === TOOL_BOSS_CHAT_PREPARE_RUN) {
+        payload = await handleBossChatPrepareRunTool({ workspaceRoot, args });
       } else if (toolName === TOOL_BOSS_CHAT_START_RUN) {
         payload = await handleBossChatStartRunTool({ workspaceRoot, args });
       } else if (toolName === TOOL_BOSS_CHAT_GET_RUN) {

package/src/test-boss-chat.js

@@ -8,6 +8,7 @@ import {
   getBossChatHealthCheck,
   getBossChatRun,
   pauseBossChatRun,
+  prepareBossChatRun,
   resumeBossChatRun,
   startBossChatRun
 } from "./boss-chat.js";
@@ -17,6 +18,7 @@ import { __testables as indexTestables } from "./index.js";
 const { handleRequest } = indexTestables;
 
 const TOOL_BOSS_CHAT_HEALTH_CHECK = "boss_chat_health_check";
+const TOOL_BOSS_CHAT_PREPARE_RUN = "prepare_boss_chat_run";
 const TOOL_BOSS_CHAT_START_RUN = "start_boss_chat_run";
 const TOOL_BOSS_CHAT_GET_RUN = "get_boss_chat_run";
 const TOOL_BOSS_CHAT_PAUSE_RUN = "pause_boss_chat_run";
@@ -202,6 +204,16 @@ async function testBossChatAdapterShouldResolveSharedConfigAndInvokeLocalCli() {
     assert.equal(health.shared_llm_config, true);
     assert.equal(health.debug_port, 9666);
 
+    const prepared = await prepareBossChatRun({
+      workspaceRoot,
+      input: {}
+    });
+    assert.equal(prepared.status, "NEED_INPUT");
+    assert.deepEqual(prepared.missing_fields, ["job", "start_from", "target_count", "criteria"]);
+    const preparedTargetQuestion = prepared.pending_questions.find((item) => item.field === "target_count");
+    assert.equal(preparedTargetQuestion.argument_name, "target_count");
+    assert.equal(prepared.next_call_example.target_count, "all");
+
     const preflight = await startBossChatRun({
       workspaceRoot,
       input: {}
@@ -301,6 +313,23 @@ async function testBossChatAdapterShouldResolveSharedConfigAndInvokeLocalCli() {
 
 async function testBossChatMcpToolsShouldValidateAndRoute() {
   await withBossChatWorkspace(async (workspaceRoot) => {
+    const toolsResponse = await handleRequest({
+      jsonrpc: "2.0",
+      id: 10,
+      method: "tools/list",
+      params: {}
+    }, workspaceRoot);
+    const tools = toolsResponse.result.tools;
+    const prepareToolSchema = tools.find((item) => item.name === TOOL_BOSS_CHAT_PREPARE_RUN).inputSchema;
+    const startToolSchema = tools.find((item) => item.name === TOOL_BOSS_CHAT_START_RUN).inputSchema;
+    assert.equal(prepareToolSchema.required, undefined);
+    assert.deepEqual(startToolSchema.required, ["job", "start_from", "target_count", "criteria"]);
+
+    const prepared = await callTool(workspaceRoot, TOOL_BOSS_CHAT_PREPARE_RUN, {}, 101);
+    assert.equal(prepared.status, "NEED_INPUT");
+    assert.deepEqual(prepared.missing_fields, ["job", "start_from", "target_count", "criteria"]);
+    assert.equal(prepared.pending_questions.find((item) => item.field === "target_count").argument_name, "target_count");
+
     const needInput = await callTool(workspaceRoot, TOOL_BOSS_CHAT_START_RUN, {}, 11);
     assert.equal(needInput.status, "NEED_INPUT");
     assert.deepEqual(needInput.required_fields, ["job", "start_from", "target_count", "criteria"]);
@@ -413,6 +442,15 @@ async function testBossChatCliShouldSupportRunAndFollowUpParsing() {
   }
 
   await withBossChatWorkspace(async (workspaceRoot) => {
+    const prepareLogs = await captureConsoleLogs(async () => {
+      await cliTestables.runBossChatCliCommand("prepare-run", {
+        "workspace-root": workspaceRoot
+      });
+    });
+    const prepared = JSON.parse(prepareLogs[0]);
+    assert.equal(prepared.status, "NEED_INPUT");
+    assert.equal(prepared.pending_questions.find((item) => item.field === "target_count").argument_name, "target_count");
+
     const logs = await captureConsoleLogs(async () => {
       await cliTestables.runBossChatCliCommand("run", {
         "workspace-root": workspaceRoot,

package/vendor/boss-chat-cli/src/cli.js

@@ -141,10 +141,15 @@ function isUnlimitedTargetCountToken(value) {
     'inf',
     'max',
     'full',
+    'allcandidates',
     '全部',
     '全量',
     '不限',
     '扫到底',
+    '全部候选人',
+    '所有候选人',
+    '全部人选',
+    '所有人选',
     '直到完成所有人选',
   ].includes(token);
 }