Firebase MCP
Gives AI development tools Firebase-specific capabilities and expertise.
Turn vague research directions into falsifiable, evidence-grounded Research Idea Briefs with a six-stage Codex workflow.
如果你是作者,可把这段代码贴到 README。徽章会链接回本页。
[](https://mcp.popnx.com/en/skills/cpt13-g-clarify-research-idea)在同一工作中把 AI 连到服务的 MCP
Gives AI development tools Firebase-specific capabilities and expertise.
同类工作,排序与 Skill 列表相同
| 本项 clarify-research-idea | timesfm | Shadowbroker | google-maps-scraper | tradingview-mcp | |
|---|---|---|---|---|---|
| 作者 | cpt13-g | google-research | BigBodyCobain | gosom | atilaahmettaner |
| Skill 数 | 1 | 1 | 1 | 1 | 1 |
| GitHub 星标 | 298 | 35,629 | 11,602 | 8,209 | 5,307 |
| 最近代码更新 | 2026年7月26日 | 2026年11月13日 | 2026年11月17日 | 2026年11月15日 | 2026年11月15日 |
| 许可证 | – | Apache-2.0 | AGPL-3.0 | MIT | MIT |
cpt13-g 提供的 1 个 Skill(指令集),AI 读取后照做 · 不是 MCP server
MCP = 手
把 AI 连到外部服务或数据,例如读取 Notion 或开发票。需要有程序或服务器在运行。
Skill = 手册
教 AI 怎么做某项工作的文本文件,不连接任何服务。
有些会重叠:少数 Skill 带有调用工具的脚本,少数 MCP 只提供知识。请从你想完成的工作出发。
Codex
依据开发者 README 和 repo 中的文件夹
GitHub account cpt13-g (person), 5 years old
★ 298 stars, few users so far · code last changed 2026年7月26日
No, plain text files the AI reads and follows
No scripts, so it does not go online by itself
No hooks or writes to machine settings found
Checked automatically from the repo files at commit 1f8334d (read only, nothing was run). Not a safety guarantee: read it and decide for yourself
第一步:你用哪个 AI?
选择你用的应用,本页只显示该应用的安装方法
This skill has no license, so we do not hand out a ZIP from our site. Download it from GitHub and arrange the folder yourself, or use "Let the AI install it" above, which is easier
Download from GitHub (whole repo) ↗
Unzip it, then zip only the .agents/skills/clarify-research-idea folder again. The ZIP must have a clarify-research-idea folder with SKILL.md at the top level, no loose files
+Create skillUpload a skillpick clarify-research-idea.zip
SettingsCapabilitiesCode execution and file creation
Turn on Code execution and file creation first (newer menus: Create and edit files) · personal plans: Settings → Capabilities · Team/Enterprise: an admin turns it on under Organization settings → Plugins & skills · every plan, including Free
Upload once; it works in claude.ai and Claude Desktop, and in recent Claude Code signed in with a Claude account (not with an API key)
Cowork and cloud sessions use the skills turned on in your claude.ai account, not ~/.claude/skills on your computer (Claude Desktop scheduled tasks run locally and still load it)
Checked against official docs 2026年11月12日 · Source: support.claude.com, support.claude.com, code.claude.com
复制这句话,粘贴到 Claude Code 的对话中
For one project, paste it into a chat that is open in that project’s folder · The AI always shows you the path before installing
The AI tells you what it will do first and asks your permission at every step
重新打开 Claude Code,输入 /clarify-research-idea,或直接提出相关任务,Claude 会自动选用
Checked against official docs 2026年11月12日 · Source: code.claude.com
Skill 是 AI 会照做的指令。安装前先阅读文件,只安装可信开发者的 Skill。
Mac:按 Command+空格输入 Terminal · Windows:下方命令在 PowerShell 中无法使用,请改用上方的“让 AI 帮你安装”
这条命令把 repo 下载到临时文件夹,只复制含 SKILL.md 的文件夹
正在检查命令 · 通过后才会显示复制按钮
ChatGPT 桌面应用内置 Codex,Codex 从 ~/.agents/skills 文件夹读取 Skill
在 ChatGPT 桌面应用中打开 Codex,把这句话粘贴到对话中
The AI always shows you the path before installing
The AI tells you what it will do first and asks your permission at every step
把 clarify-research-idea 文件夹放到 ~/.agents/skills/clarify-research-idea
~ 是你的主文件夹($HOME)
Skill 是 AI 会照做的指令。安装前先阅读文件,只安装可信开发者的 Skill。
Mac:按 Command+空格输入 Terminal · Windows:下方命令在 PowerShell 中无法使用,请改用上方的“让 AI 帮你安装”
这条命令把 repo 下载到临时文件夹,只复制含 SKILL.md 的文件夹
正在检查命令 · 通过后才会显示复制按钮
Checked against official docs 2026年11月15日 · Source: learn.chatgpt.com
不能使用你自己安装的 Skill,只能使用目录中插件自带的 Skill
想在 ChatGPT 中使用这个 Skill,请按上面的方法使用 ChatGPT 桌面应用
Checked against official docs 2026年11月15日 · Source: learn.chatgpt.com
复制这句话,粘贴到 Codex 的对话中 · 适用于 ChatGPT 桌面应用中的 Codex、Codex CLI 和 IDE 扩展
For one project, paste it into a chat that is open in that project’s folder · The AI always shows you the path before installing
The AI tells you what it will do first and asks your permission at every step
Codex 会自动发现新安装的 Skill;如果没有出现,请重启 Codex · 使用方法:在 ChatGPT 中输入 @ 并选择 Skill · 在 Codex CLI 或 IDE 扩展中运行 /skills,或输入 $ 加 Skill 名称
Checked against official docs 2026年11月15日 · Source: learn.chatgpt.com
Skill 是 AI 会照做的指令。安装前先阅读文件,只安装可信开发者的 Skill。
Mac:按 Command+空格输入 Terminal · Windows:下方命令在 PowerShell 中无法使用,请改用上方的“让 AI 帮你安装”
这条命令把 repo 下载到临时文件夹,只复制含 SKILL.md 的文件夹
正在检查命令 · 通过后才会显示复制按钮
复制这句话,粘贴到 Cursor 的对话中
For one project, paste it into a chat that is open in that project’s folder · The AI always shows you the path before installing
The AI tells you what it will do first and asks your permission at every step
在 Cursor 的 Agent 聊天中输入 / 并选择 clarify-research-idea
Checked against official docs 2026年11月12日 · Source: cursor.com
Skill 是 AI 会照做的指令。安装前先阅读文件,只安装可信开发者的 Skill。
Mac:按 Command+空格输入 Terminal · Windows:下方命令在 PowerShell 中无法使用,请改用上方的“让 AI 帮你安装”
这条命令把 repo 下载到临时文件夹,只复制含 SKILL.md 的文件夹
正在检查命令 · 通过后才会显示复制按钮
适用于其他能读取 Skill 的应用。这句话列出了 Claude Code、Codex 和 Cursor 的路径,AI 会按你的应用选择 · Copy this sentence into the app’s chat
For one project, paste it into a chat that is open in that project’s folder · The AI always shows you the path before installing
The AI tells you what it will do first and asks your permission at every step
Checked against official docs 2026年11月12日 · Source: code.claude.com, learn.chatgpt.com, cursor.com
其他支持 Agent Skills 的应用(如 Gemini CLI、GitHub Copilot)使用不同文件夹,请查看该应用的指南,再把含 SKILL.md 的文件夹复制过去。
已于 2026年11月11日 对照 Claude Code、Codex 和 Cursor 官方指南核实 skill 文件夹
以下文字来自开发者的 repo,不是 PopNx MCP 的指令
显示前 8,000 个字符 · 在 GitHub 打开完整文件
---
name: clarify-research-idea
description: Turn vague or partially formed AI, computer-science, data-driven, and engineering-method research directions into evidence-grounded, falsifiable, paper-ready Research Idea Briefs through heuristic questioning, staged confirmation, literature positioning, novelty attack, and method-figure planning. Use when the user needs to clarify a research idea, formalize inputs/outputs/objectives/constraints, frame background and motivation, select 3-5 anchor papers, identify a defensible gap, derive 2-3 contributions, stress-test novelty, or design an input-to-output method framework and core-paper figure. Also trigger on requests such as 明确科研 idea、启发式梳理 idea、互动追问科研问题、梳理论文创新点、定义研究任务、找对标论文、把研究方向变成论文方案、画方法框架图. Do not use as the primary workflow for running experiments, exhaustive systematic reviews, drafting full manuscript sections, or polishing finished prose.
---
# Clarify Research Idea
Convert an early research direction into an auditable idea architecture before detailed experiment design or manuscript writing. Preserve the researcher's judgment: make assumptions and framing visible, challenge weak novelty, and never manufacture supporting evidence.
## Choose the operating mode
Use `guided` by default when the question, gap, or method is uncertain. Ask one high-information question at a time, maintain a visible working draft, and pause at the three checkpoints.
Use `synthesize` when the user supplies enough material or explicitly requests a one-pass result. Complete the brief with labeled assumptions; mark checkpoint-dependent claims `UNCONFIRMED` and request confirmation at the end.
State the selected mode, research domain, maturity level, and intended deliverable in one short line. Let the user correct them cheaply.
## Run the heuristic interaction loop
Read [interactive-heuristics.md](references/interactive-heuristics.md) in `guided` mode, whenever the user's idea is vague, or whenever the user asks for interactive clarification. Treat interaction as part of the method, not as casual clarification.
Use this turn pattern:
1. State the current bottleneck in one sentence.
2. Ask one high-information question that will change the idea architecture.
3. Offer 2-3 plausible answer directions when the user is unsure.
4. Update the visible `Working Idea Canvas` with provenance labels before asking the next question.
Do not overwhelm the user with a survey. If multiple questions are useful, choose the one that best constrains the task, gap, hypothesis, or evidence path. If the user asks for a fast one-pass result, switch to `synthesize` and mark unconfirmed decisions explicitly.
## Maintain provenance
Label substantive content throughout the workflow:
- `PROVIDED`: stated or supplied by the user.
- `VERIFIED`: supported by a checked primary source or supplied artifact.
- `INFERRED`: reasoned from available material but not directly established.
- `PROVISIONAL`: useful working hypothesis awaiting evidence or user confirmation.
- `UNKNOWN`: missing and consequential.
Do not silently upgrade one status to another. When browsing or academic-search tools are available, verify time-sensitive literature claims and bibliographic metadata. Treat failure to find a paper as search incompleteness, not proof of novelty.
## Run the workflow
### 0. Scope the session
Determine the application, research object, current idea, intended venue or audience, available data or system, resource limits, and desired output language. Stay centered on AI, computer science, data-driven science, and engineering-method papers. If the request is primarily qualitative or non-methodological, explain that the mathematical contract must be adapted before continuing.
Identify whether the user has: a broad interest, a question without a task definition, a method without positioning, anchor papers without a gap, or a nearly complete idea needing audit. Use the maturity triage in [interactive-heuristics.md](references/interactive-heuristics.md) to choose the first question. Reuse supplied documents and prior decisions instead of re-asking settled questions.
### 1. Establish the Task Contract
Read [task-formalization.md](references/task-formalization.md). Define the inference-time inputs, required outputs, data or interaction process, model or decision function, objective, constraints, evaluation measures, baselines, assumptions, and explicit exclusions. Use mathematical notation only when it removes ambiguity; do not invent a loss function for appearance.
Require alignment among the claimed problem, information available at use time, optimization target, and evaluation metric.
**Checkpoint 1 — Task Contract:** Present the one-sentence task, compact formulation, success criterion, assumptions, and scope boundaries. In `guided` mode, obtain user confirmation before treating the contract as fixed.
### 2. Make the argument and framing explicit
Read [framing-and-motivation.md](references/framing-and-motivation.md). Build the chain:
`background -> consequential friction -> unresolved cause -> required capability -> proposed research move`
Record the intended reader update, the current default view, and the strongest counter-position in a `Framing Intent` block. Allow framing to guide emphasis and argument order, but never let it bias evidence retrieval, omit material counterevidence, or turn a hypothesis into a fact.
Express each challenge as a causal requirement, not an adjective: property of the setting -> technical difficulty -> why generic methods fail -> capability the solution must provide.
### 3. Position against 3-5 anchor papers
Read [anchor-paper-positioning.md](references/anchor-paper-positioning.md). Select papers by role rather than fame: problem anchor, method anchor, closest competitor, experiment anchor, and optional writing/architecture anchor. A paper may fill more than one role.
Verify title, authors, year, venue or repository, and DOI or stable URL. Extract `WHY / HOW / WHAT / LIMIT / TRANSFER` for each paper. Compare task assumptions and evidence, not only module names. Derive high-level writing characteristics without copying distinctive wording.
### 4. Derive and attack the idea
Read [novelty-and-falsification.md](references/novelty-and-falsification.md). Consider three routes: a genuinely new problem, an old problem with an ignored condition or failure mode, or new evidence/understanding that changes a boundary. Force each candidate through:
`observed gap -> underlying cause -> design principle -> method component -> expected evidence`
Generate alternatives only when they help the user choose. Reduce the selected idea to one core hypothesis and 2-3 contribution cards. Attack it with the closest existing method, the simplest plausible baseline, alternative explanations, confounds, negative cases, and feasibility limits. In `guided` mode, run the reviewer-simulation questions from [interactive-heuristics.md](references/interactive-heuristics.md) before treating a contribution as stable.
**Checkpoint 2 — Positioning and Hypothesis:** Present the verified anchors, defensible gap, rejected alternatives, core hypothesis, and falsifier. In `guided` mode, obtain user confirmation before freezing the method architecture.
### 5. Design the method architecture and core figure
Read [framework-figure.md](references/framework-figure.md). Map every challenge to a design principle, every principle to a component or procedure, and every component to a claim and validation hook. Build an editable textual specification and a Mermaid draft using:
`context/input -> problem structure -> method components -> objective/constraints -> output -> evidence`
Keep the figure causal and selective. Do not add decorative modules merely to make the method look complex. If the user requests an actual `.pptx`, hand the approved figure specification to a presentation-generation workflow rather than making PowerPoint a hidden dependency of 提问、分享技巧或报告无法使用。按“有帮助”数量排序。