Research

v1.0.0

在开始任何分析报告前使用。通过结构化对话理清研究目标、范围、受众和关键问题,形成研究大纲。

0· 111·0 current·0 all-time

Install

OpenClaw Prompt Flow

Install with OpenClaw

Best for remote or guided setup. Copy the exact prompt, then paste it into OpenClaw for mingginwan/data-research.

Previewing Install & Setup.
Prompt PreviewInstall & Setup
Install the skill "Research" (mingginwan/data-research) from ClawHub.
Skill page: https://clawhub.ai/mingginwan/data-research
Keep the work scoped to this skill only.
After install, inspect the skill metadata and help me finish setup.
Use only the metadata you can verify from ClawHub; do not invent missing requirements.
Ask before making any broader environment changes.

Command Line

CLI Commands

Use the direct CLI path if you want to install manually and keep every step visible.

OpenClaw CLI

Bare skill slug

openclaw skills install data-research

ClawHub CLI

Package manager switcher

npx clawhub@latest install data-research
Security Scan
VirusTotalVirusTotal
Pending
View report →
OpenClawOpenClaw
Benign
high confidence
Purpose & Capability
Name and description promise a structured research kickoff; the skill is instruction-only and asks the agent to elicit scope, audience, questions, propose research paths, produce and save a research outline — all coherent with a 'research' planning assistant. It does not request unrelated binaries, credentials, or external services.
Instruction Scope
SKILL.md stays on scope: it guides a dialogue to clarify goals, scope, audience, proposes 2-3 paths, presents an outline, saves it, and waits for explicit user confirmation before proceeding (a hard gate). It asks to check user-provided files/links/data (expected) but does not instruct reading system files, environment variables, or contacting unexpected endpoints. It explicitly limits subsequent calls to the 'report-outline' skill only.
Install Mechanism
No install spec and no code files — instruction-only. Nothing is downloaded or written by an installer, which minimizes disk/execute risk.
Credentials
No required env vars, credentials, or config paths. The only data access implied is user-provided materials (files/links/data) which are appropriate for producing a research outline.
Persistence & Privilege
The skill is not always-enabled and uses the platform default model invocation behavior. It does instruct saving an outline file (expected) but does not request persistent privileges, modify other skills, or demand elevated system access.
Assessment
This skill is coherent and low-risk: it guides structured planning and refuses to start analysis until the user confirms the outline. Before installing, consider: (1) where the saved outline will be stored and whether that storage location is acceptable, (2) avoid providing sensitive credentials or secrets in the files/links you ask the skill to inspect, (3) verify that the downstream 'report-outline' skill (which this skill will call) is trusted before proceeding, and (4) be aware that the agent will write the confirmed outline to disk under the specified filename pattern. If you need the skill to access protected data or specific storage, verify those details first.

Like a lobster shell, security has layers — review code before you run it.

latestvk975bwdrdmfj5rymm2g7zsgc6n83qf68
111downloads
0stars
1versions
Updated 1mo ago
v1.0.0
MIT-0

结构化研究启动

通过自然对话,将模糊的分析需求转化为清晰的研究大纲。

先理解背景和目的,再逐步确认范围、受众、关键问题,最终输出研究大纲供确认。

<HARD-GATE> 在研究大纲获得用户确认前,不要开始写报告、做分析、或产出任何结论。无论需求看起来多简单,都必须先过这个流程。 </HARD-GATE>

反模式:"这个分析很简单不需要规划"

每个分析项目都走这个流程。一个简单的对比分析、一个数据汇总、一个趋势判断——都需要。"简单"的任务恰恰最容易因为未审视的假设而浪费时间。大纲可以很短(简单任务几句话即可),但必须呈现并获得确认。

检查清单

按顺序完成以下步骤:

  1. 了解背景 — 检查相关文件、数据源、已有资料
  2. 澄清问题 — 逐个提问,理解目的、约束、成功标准
  3. 确认受众 — 报告给谁看?他们的知识水平和关注点是什么?
  4. 提出 2-3 种研究路径 — 带权衡分析和推荐
  5. 呈现研究大纲 — 按模块逐步确认
  6. 保存研究大纲 — 写入文件并确认
  7. 大纲自查 — 检查占位符、矛盾、模糊点、范围
  8. 用户审阅大纲 — 等待用户确认后才继续
  9. 进入下一步 — 调用 report-outline skill 制定报告结构

流程

了解背景 → 逐个提问澄清 → 确认受众 → 提出研究路径
→ 呈现大纲 → 用户确认? → (否:修改) / (是:保存)
→ 大纲自查 → 用户审阅 → 调用 report-outline

详细步骤

理解需求:

  • 先查看用户提供的相关资料(文件、链接、数据)
  • 评估范围:如果需求涉及多个独立主题(如"分析市场、竞品、技术趋势和财务状况"),先标记需要拆分,不要在细节上花时间
  • 如果项目太大,帮助用户拆分为子课题:有哪些独立部分?它们之间什么关系?先从哪个开始?每个子课题走独立的 研究 → 大纲 → 撰写 循环
  • 对于范围合适的项目,逐个提问深入理解
  • 尽量用选择题,开放题也可以
  • 每条消息只问一个问题
  • 聚焦于:目的、约束、成功标准、受众

探索路径:

  • 提出 2-3 种不同的研究路径,带权衡分析
  • 用对话方式呈现,给出推荐及理由
  • 先说推荐方案,再解释为什么

呈现研究大纲:

  • 一旦你认为理解了需求,呈现研究大纲
  • 每个模块根据复杂度调整篇幅:简单的几句话,复杂的 200-300 字
  • 每个模块确认后再继续
  • 覆盖:研究问题、数据来源、分析方法、预期产出
  • 随时可以回头澄清

大纲之后

保存文档:

  • 将确认的研究大纲保存到合适的位置
  • 文件名格式:YYYY-MM-DD-<主题>-研究大纲.md

大纲自查:

  1. 占位符扫描: 有没有"待定"、"TODO"、未完成的部分?修掉。
  2. 内部一致性: 各部分有没有矛盾?分析方法和研究问题匹配吗?
  3. 范围检查: 这个范围适合一份报告吗?还是需要拆分?
  4. 模糊检查: 有没有可以被两种方式理解的需求?如果有,选一种并明确。

发现问题直接修改,不需要重新审阅。

用户审阅:

"研究大纲已保存到 <路径>。请审阅,有修改意见告诉我,确认后我们开始制定报告结构。"

等待用户回复。有修改则修改后重新自查。确认后才继续。

下一步:

  • 调用 report-outline skill 制定报告结构
  • 不要调用其他 skill

核心原则

  • 每次只问一个问题 — 不要连珠炮
  • 选择题优先 — 比开放题更容易回答
  • YAGNI — 砍掉不必要的分析范围
  • 探索多条路径 — 先提 2-3 种方案再定
  • 逐步确认 — 呈现大纲,确认后才继续
  • 保持灵活 — 发现不清楚的随时回头澄清

Comments

Loading comments...