Nagi Skills
A skill collection for questioning ideas, comparing implementation options and checking an agent’s work with evidence
- 526 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- Some skills ask the agent to run checks and edit interface files.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/nagi-skillsDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Some skills ask the agent to run checks and edit interface files.
This is third-party code. Review the repository files before installing.
What it does
Nagi Skills gives agents several ways to examine an idea and its implementation. A questioning skill works in rounds until important unknowns become explicit. An architecture skill then compares options based on the credentials, permissions and services the user can manage. Verification calls for tests and interaction in a real browser. A separate skill reviews interface screenshots and removes repetitive visual choices that do not fit the page.
Who it is for. For people learning to work with coding agents who want to question decisions and check evidence.
Good fit when
- Question an idea before implementation
- Choose between technical approaches
- Ask for browser evidence after an agent reports completion
- Review a finished interface from screenshots
Not a fit when
- You need one fixed architecture for every project
- You want implementation without discussing requirements
- You need a specialist security auditor
Example request
Question my service idea. Stop when the remaining unknowns no longer affect the next step.Limitations
The main material is Chinese, so request another response language explicitly. Interface verification needs browser tools in the host agent; installing the collection does not provide them. Installed skills do not automatically follow repository updates and need a separate update. The `-g` option installs at user level, making the skills available across that agent’s projects until removed or replaced with a project-scoped installation.
How to disable. Remove the collection’s skills from the selected agent’s global directory using your skill manager.
FAQ
Can I change questioning intensity?
Yes. The `grill` skill offers five intensity levels.
Can I install only one skill?
Yes. Select the relevant skill during installation.
How do I select an agent?
Use the installer’s `-a <agent>` option with your client name.
Is there a plugin option?
Yes. Claude Code can add the repository as a marketplace and install `nagi-skills@nagi-studio`.
How do I activate newly installed skills?
Restart the session so the agent reloads its skill list.
Discussion
No questions or reviews yet.
Related
SkillsA catalog of nearly two hundred skills: code delivery, review, debugging, AI governance, product work and agent coordination.
A skill that designs, tests, evaluates and packages other agent skills, starting from architecture
A skill registry with a CLI and API: the agent describes a task and gets a recommendation with a trust score and install command
Internet Court
Internet Court Skill
A skill for agent-to-agent deals: one entry point to identity, negotiation, contracts, payments, escrow and dispute resolution
A completion-discipline skill for large agent tasks: write the acceptance ledger with runnable checks first, then do the work, then reverify against evidence
A tool that builds tested skills from a real workflow and governs their lifecycle through your own marketplace
README in short
The collection accompanies a video series about working with AI. Claude Code can install it as a plugin; the general installer supports agent selection and project or global scope.
README badge
Original SKILL.md
--- name: grill description: 分轮追问用户的计划、决定或想法，直到重要未知暴露出来，再动手执行。适用于任何领域，不限于写代码。当用户想理清思路、给想法做压力测试，或者说出这些话时使用：拷打我、追问我、帮我想清楚、帮我理一理、挑挑毛病、有什么没想到的、grill me、stress-test this、poke holes in this、interview me。 --- 先问拷打强度，以及要不要留决策记录（decision log）。用户已说明的不再问。 1. 浅问：只补最关键的缺口 2. 常规：把明显没想清楚的地方问明白 3. 细问：检查目标、约束、取舍和默认假设 4. 深挖：主动寻找风险、遗漏和边界情况 5. 拷打：严格挑战前提，直到重要未知基本暴露 用户可以随时调整强度。 要留记录时，每定一条就追加「结论 + 理由」，另列待定问题。能写文件就新建 `docs/decisions-<英文主题>.md`；不能就每隔几轮贴出完整版供复制。 分轮进行。每轮询问当前已经具备前提的问题，可以是一条，也可以是多条。依赖尚未确定答案的问题，留到下一轮。 优先追问会明显影响后续方案的内容： - 真正想解决什么问题 - 哪些事情已经确定 - 哪些事情还没有决定 - 哪些前提只是用户默认成立 - 有没有用户尚未意识到的重要问题 - 哪些取舍会改变后续方向 能自己查清楚的事实不要交给用户；涉及用户的目标、偏好、成本和取舍时，再让用户决定。必要时提供选项和建议，但不要替用户决定。 每个问题使用： ``` ❓ Q1｜问题标题 问题内容。需要时给出选项。 ➡️ 建议：简短给出你的推荐和理由。 ``` 同一轮有多个问题时继续 Q2、Q3。问题保持具体、容易回答。不要一次堆出大量低价值问题。