cue-research-cli
Cue 深度研究 CLI(@cueai/cue-cli,命令 cue)使用说明。用于行业/公司/人物/政策/合规类深度调研:一句话发起多智能体研究,输出带来源的 Markdown 报告,可落盘、可续接会话追问、可指定搭子模板。涉及「深度调研」「尽调」「行研」「公司研究」「竞品对比」「政策跟踪」「人物核查」等需要多源检…
Skill by sensedeal
Chalk Labs
Need help deploying cue-research-cli?
Chalk Labs sets up agents like this inside real teams: connected to your tools, tested on your data, with a human in the loop where it matters.
Install
Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.
npx agentmarket install skill/sensedeal-cue-research-cli
Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.
Get the weekly diff
The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.
About
Use cases: documents, research
Reviews
Write a reviewNo reviews yet. Installed it? Reviews come from verified installers only. How reviews work
Permissions
What this skill can do and what it asks for, from its SKILL.md and code. Checked Oct 25, 2026.
- Tools
- No tools found
- Secrets
- None asked for
- Install scripts
- None found
Network hosts it talks to 1
codeload.github.com· found in install command
Files it touches
~/.claude/skills/cue-research-cli
What's inside
- Skill name
cue-research-cli- Description
- Cue 深度研究 CLI(@cueai/cue-cli,命令 cue)使用说明。用于行业/公司/人物/政策/合规类深度调研:一句话发起多智能体研究,输出带来源的 Markdown 报告,可落盘、可续接会话追问、可指定搭子模板。涉及「深度调研」「尽调」「行研」「公司研究」「竞品对比」「政策跟踪」「人物核查」等需要多源检索与交叉验证的场景时使用。
- License
- MIT
- Instructions
- 98 lines in SKILL.md · view on GitHub
Files (1)
- SKILL.md
SKILL.md (first lines)
# Cue 调研搭子(cue CLI)
`cue`(npm 包 `@cueai/cue-cli`)是 Cue(cuecue.cn)多智能体深度研究的本地命令行桥梁。WorkBuddy 负责安装与登录态调度,本 Skill 只负责「拿到命令后怎么调」。
**核心命令:`cue research`** —— 把一句自然语言问题交给 Cue,自动拆解子任务、多源检索交叉验证,返回带来源引用的 Markdown 研究报告。
> ⚠️ 本 skill 与仓库里那个**同名的 Python 形态** `cue-research/` 是两件事:那个是
> 让 agent 直接跑 Python 脚本的形态,本目录是 **CLI 形态**(命令 `cue`,由 WorkBuddy 安装)。
> 两者名字相近但入口不同,别混用。
## 一、何时使用 / 何时不用
**应使用**:需要多源检索 + 交叉验证 + 出报告的开放性问题,例如行业研究、公司/竞品调研、政策与监管跟踪、人物背景核查、合规风险体检、投前/贷前尽调。
**不应使用**:
- 单条事实查询(某股票现价、今天天气)—— 直接用行情/搜索工具,不要为它跑一次深度研究;
- 纯本地文件处理(解析 PDF、改写文档)—— Cue 不做这个(`cue` 没有文档解析命令);
- 需要秒级返回的交互 —— 一次研究通常数分钟。
## 二、前置条件
1. CLI 由 WorkBuddy 通过 `npm install -g @cueai/cue-cli` 安装(Node ≥ 20.12,由 WorkBuddy 托管运行时);
2. 登录态:`cue auth status`(Windows 为 `cue.cmd auth status`)输出「已登录」才算可用,退出码 0=已登录 / 2=未登录;
3. 未登录时引导用户在连接器面板点「连接」:`cue auth login` 会打印 https:// 授权 URL,由 WorkBuddy 打开浏览器完成设备码授权;
4. **浏览器授权暂不可用时的退路**:若 `cue auth login` 明确提示「服务端尚未开放设备码授权」,改用用户自填 Key —— 到 https://cuecue.cn/api-key 创建 key,然后 `cue auth login --api-key <key>`,或设 `export CUE_API_KEY=sk-...` 后重跑。**看到该提示时直接走这条退路,不要反复重试 `auth login`**;
5. 排障先看 `cue doctor`(输出 JSON:node 版本、配置文件、脱敏 key、API 连通性)。
> **登录态是怎么维持的**:设备码换来的访问令牌 1 小时过期,但 CLI 会在任何 API 调用
> 遇到 401 时**自动静默刷新一次并重试**。所以**不要因为一次 401 就急着让用户重新登录** ——
> 只有刷新也失败时,CLI 才会以退出码 2 结束,那时再引导重连。
## 三、命令速查
| 命令 | 用途 |
|---|---|
| `cue research "<问题>"` | 发起一次研究,报告打到 stdout(前台阻塞到完成) |
| `cue research "<问题>" -o <文件>` | 报告落盘为 Markdown,并附来源清单 |
| `cue research "<问题>" --json` | 输出 `{conversationId, elapsedMs, report, sources, timeline}` |
| `cue research "<问题>" --conversation-id <id>` | 续接同一会话追问 |
| `cue research "<问题>" --template-id <id>` | 用指定搭子模板跑 |
| `cue playbook list` / `cue playbook get <id>` | 查看搭子模板(列表 / 详情) |
| `cue doctor` | 环境与连通性自检(JSON,无副作用) |
| `cue auth login` / `status` / `logout` | 授权 / 查登录态 / 登出清凭证 |
Windows 下命令名用 `cue.cmd`,参数完全一致。
## 四、`cue research` 参数
| 参数 | 类型 | 必填 | 默认 | 说明 |
|---|---|---|---|---|
| `<query>` | string | 是 | — | 研究问题,自然语言,用引号包裹 |
| `--template-id <id>` | string | 否 | — | 搭子模板;裸 id 自动补 `template_` 前缀,纯数字后缀直接报错 |
| `--conversation-id <id>` | string | 否 | 随机 | 续接会话;ID 来自上次输出的 `conversation_id` |
| `-o, --output <file>` | string | 否 | — | 落盘 Markdown,支持 `~/`,目录自动创建 |
| `--timeout <seconds>` | number | 否 | 1200 | 整轮超时 |
| `--json` | flag | 否 | false | 结构化输出,适合再做二次处理 |
| `-q, --quiet` | flag | 否 | false | 不打印进度(进度默认写 stderr) |
## 五、推荐调用姿势
**落盘(默认用法)**:
```bash
cue research "比亚迪 DM-i 与长城 Hi4 混动技术路线差异" -o reports/byd-hi4.md
```
**要结构化结果**(需同时拿到来源与时间线):加 `--json`。
**续接追问**:用上次输出的 `conversation_id` 加 `--conversation-id`,在原研究上下文里追问(如「再补一节供应商集中度」)。
**用搭子模板**:先 `cue playbook list` 找到 `template_id`,再 `--template-id` 传入。
## 六、耗时与额度
- 单次研究通常数分钟,默认超时 1200 秒;
- 一次只跑一个研究,不要并发;
- 深度研究消耗 Cue 额度,问题越聚焦越省;不要为验证连通性反复提交真实研究(用 `cue doctor` 验证连通即可)。
## 七、错误处理
| 退出码 | 含义 | 处理 |
|---|---|---|
| 0 | 成功 | — |
| 2 | 缺少/失效凭证(含 401、403) | 重新连接;`cue auth login` |
| 3 | 入参错误(如纯数字 template_id) | 改参数后重试 |
| 4 | 网络不可达 | 检查网络/代理/`--base-url` |
| 5 | 超时 | 加大 `--timeout`,或用 `--conversation-id` 续跑 |
| 1 | 其他错误 | 读 stderr 原文 |
「未捕获到报告正文」时 stderr 会给出 `kind=...`(`no_agent_events` / `stream_cut_before_reporter` / `reporter_started_no_text`)与 `conversation_id`,据此续跑或调超时,不要凭空补写内容。
## 八、纪律
- 不在命令行、日志或回复里输出 API Key;非必要不用 `--api-key`(会进进程列表);
- 报告结论以 Cue 返回的正文与来源为准,缺失部分标注「未检索到」,**不要补写未经验证的数据**;
- 需要用户决策的高风险动作(对外发送、下单、删除)先确认;研究本身只读取公开信息与用户自有数据。From sensedeal/cue-skills, MIT.
More from sensedeal/cue-skills
- 0.627score
cue-research
sensedeal
Use when the user asks a research question they want Cue to run — against a saved 搭子(buddy) template or as free-form deep research.
SkillAFreshness: A★ 6Claude Code +9 - 0.627score
cue-data-mcp
sensedeal
Use when the user wants an external AI agent to query Cue's public MCP data services — regulatory, macro, disclosures, market data, legal full text, holdings…
SkillAFreshness: A★ 6Claude Code +9 - 0.627score
cue-omni-reader
sensedeal
Use when the user wants an external AI agent to parse or understand an HTTP(S) URL or an authorized local document, audio, or video source through Cue Omni…
SkillAFreshness: A★ 6Claude Code +9 - 0.619score
cue-deep-verification
sensedeal
用 Cue 跑「深度核查」场景的深度研究:多源公开数据交叉、结论带来源。 Run Cue deep research for the "Deep Verification" scenario.
SkillAFreshness: A★ 6Claude Code +9 - 0.619score
cue-equity-research
sensedeal
用 Cue 跑「投资研究」场景的深度研究:多源公开数据交叉、结论带来源。 Run Cue deep research for the "Equity Research" scenario.
SkillAFreshness: A★ 6Claude Code +9 - 0.619score
cue-financial-deep-read
sensedeal
用 Cue 跑「财报深读」场景的深度研究:多源公开数据交叉、结论带来源。 Run Cue deep research for the "Financial Deep Read" scenario.
SkillAFreshness: A★ 6Claude Code +9 - 0.619score
cue-fund-advisory
sensedeal
用 Cue 跑「基金投顾」场景的深度研究:多源公开数据交叉、结论带来源。 Run Cue deep research for the "Fund Advisory" scenario.
SkillAFreshness: A★ 6Claude Code +9 - 0.619score
cue-global-macro
sensedeal
用 Cue 跑「全球宏观」场景的深度研究:多源公开数据交叉、结论带来源。 Run Cue deep research for the "Global Macro" scenario.
SkillAFreshness: A★ 6Claude Code +9 - 0.619score
cue-industry-research
sensedeal
用 Cue 跑「行业研究」场景的深度研究:多源公开数据交叉、结论带来源。 Run Cue deep research for the "Industry Research" scenario.
SkillAFreshness: A★ 6Claude Code +9 - 0.619score
cue-insurance-marketing
sensedeal
用 Cue 跑「保险营销」场景的深度研究:多源公开数据交叉、结论带来源。 Run Cue deep research for the "Insurance Marketing" scenario.
SkillAFreshness: A★ 6Claude Code +9 - 0.619score
cue-ipo-research
sensedeal
用 Cue 跑「IPO研究」场景的深度研究:多源公开数据交叉、结论带来源。 Run Cue deep research for the "IPO Research" scenario.
SkillAFreshness: A★ 6Claude Code +9 - 0.619score
cue-legal-compliance
sensedeal
用 Cue 跑「法律合规」场景的深度研究:多源公开数据交叉、结论带来源。 Run Cue deep research for the "Legal & Compliance" scenario.
SkillAFreshness: A★ 6Claude Code +9
Signals
- Stars
- 6
- Forks
- 0
- Open issues
- 0
- Last commit (this folder)
- Oct 9, 2026 (22 days ago)
- Last release
- Oct 9, 2026 (21 days ago)
From GitHub (fetched Oct 24, 2026).
Details
- License
- MIT
- Repository
- github.com/sensedeal/cue-skills
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
- Source
- GitHubgithub-graphql
More from sensedeal
- 0.627score
cue-omni-reader
sensedeal
Use when the user wants an external AI agent to parse or understand an HTTP(S) URL or an authorized local document, audio, or video source through Cue Omni…
SkillAFreshness: A★ 6Claude Code +9 - 0.627score
cue-data-mcp
sensedeal
Use when the user wants an external AI agent to query Cue's public MCP data services — regulatory, macro, disclosures, market data, legal full text, holdings…
SkillAFreshness: A★ 6Claude Code +9 - 0.627score
cue-research
sensedeal
Use when the user asks a research question they want Cue to run — against a saved 搭子(buddy) template or as free-form deep research.
SkillAFreshness: A★ 6Claude Code +9 - 0.619score
cue-cross-border-legal
sensedeal
用 Cue 跑「涉外法律」场景的深度研究:多源公开数据交叉、结论带来源。 Run Cue deep research for the "Cross-border Legal" scenario.
SkillAFreshness: A★ 6Claude Code +9 - 0.619score
cue-credit-diligence
sensedeal
用 Cue 跑「信贷尽调」场景的深度研究:多源公开数据交叉、结论带来源。 Run Cue deep research for the "Credit Diligence" scenario.
SkillAFreshness: A★ 6Claude Code +9 - 0.619score
cue-deep-verification
sensedeal
用 Cue 跑「深度核查」场景的深度研究:多源公开数据交叉、结论带来源。 Run Cue deep research for the "Deep Verification" scenario.
SkillAFreshness: A★ 6Claude Code +9
Done for you
Rather not set it up yourself?
Chalk Labs scopes, builds and runs agents for your team: the task audit, the eval harness, the integrations and the weekly ops report. Fixed scope, priced before we start.
See how we build AI agents →Read the case studies
Chalk Labs is a two-founder AI and growth studio. We build AI agents, CRMs and MVPs for Web3 and AI companies, and have shipped 50+ product and token launches.
- launches shipped
- 50+
- launches shipped
- media placements
- 500+
- media placements
- founders on every account
- 2
- founders on every account
Press we've placed: Forbes · CoinDesk · Decrypt · Cointelegraph · The Block
Rahil Jain, Co-founder, Chalk Labs. Ex-Goldman Sachs engineer. You talk to the person who builds it.
Get cue-research-cli running in your business
Tell us what you want to automate. A founder replies within one working day with a plan, not a pitch deck.
Cite this page
cue-research-cli skill. Agent Mart. https://agentmart.dev/skills/sensedeal-cue-research-cli (last updated 2026-10-24)