L
Lusona/ai-skills
Iterative product and engineering work with explicit outcomes, evidence, authorization, TDD, and bounded…
在 GitHub 查看 ↗链接已于 2026-11-12 核查 · HTTP 200 · 打开的是厂商自己的官网
安装与配置
客户端配置（取自该仓库的安装命令）
{
"mcpServers": {
"lusona-ai-skills": {
"command": "npx",
"args": ["skills@latest"]
}
}
}解析自我们于 2026-11-08 抓取的该仓库 README；我们未执行过。
command/args 来自仓库写明的安装命令；不确定时请以该项目 README 为准。
VS Code（.vscode/mcp.json —— 键名是 servers）
{
"servers": {
"lusona-ai-skills": {
"command": "npx",
"args": ["skills@latest"]
}
}
}VS Code 顶层键是 <code>servers</code>，其余与上面那份相同。
Claude Code（命令行）
claude mcp add --transport stdio lusona-ai-skills -- npx skills@latest
Claude Code 不收 JSON 文件，用这条命令直接加。
配置项的键名以你的客户端为准：客户端与 CLI
- ★ 数取值
- 3 ★ · Lusona/ai-skills · 取自 GitHub 搜索 API · （2026-11-13 20:00:50 核对）
- 来源明细
- 从该仓库的 GitHub 页面直接读取 · 条目类型：开源项目
- 收录与活跃
- 收录于 2026-11-09 · 最近见到 2026-11-14 · 共见到 6 次 · 被 1 个来源列出
- 评分构成
- 综合分 103 · 3 ★ · 1 个来源 · repo
- 安装
npx skills@latest
常见问题（答案都来自我们的记录）
怎么安装？
该仓库给出的安装命令是
npx skills@latest。我们的记录里它是仓库自己提供的；不保证在你的客户端上直接可用。它适合谁 / 用来做什么？
我们把它归在这些标签下: Agent 技能。标签来自条目自身，不是我们强加的分类。
谁维护它、还在更新吗？
仓库
Lusona/ai-skills；我们收录于 2026-11-09；最近见到 2026-11-14。这里没有"官方认证"或"质量评级"——我们只记录来源与时间。# Lusona Skills
以下摘录自该仓库的 README（片段） · 在 GitHub 读完整 README →
English | Español | Deutsch | 简体中文 | 日本語
Lusona organizes coding-agent skills around a shared work loop:
This is a common interaction loop. Start where needed and revisit steps as evidence changes.
Choose, act, check, repeat. Choose the next action from the outcome and current
evidence. Revisit exploration, investigation, design, implementation, or review as needed.
There is no fixed skill sequence.
Revise decisions as you learn. When evidence challenges an earlier decision, revisit
it within the agreed authority. Update affected records so subsequent work follows the
revised decision.
Continue within authorization. Resolve routine choices and keep working without
repeated approval. Ask when a decision exceeds existing authority.
Verify completion. Complete the agreed scope, required checks and review, and any
required acceptance or delivery. Report what remains when completion is blocked.
For Claude Code, Codex, Cursor, Xcode 27 coding agents, and Google Antigravity App, CLI, and IDE · Checked-out source version 2.0.1
> [!WARNING]
> Do not install Lusona using Vercel Skills (npx skills@latest); use the native plugin installation.
Read why we built Lusona Skills: Why we built Lusona Skills.
This README describes the source in this checkout. GitHub installation uses committed,
pushed sources; local changes do not establish a published release.
## Index
Where to start
Choose a command
Installation
How the work continues
Questions and evidence
Recovery and resumption
Results and handoffs
Reference
Contributing
License
Development
## Where to start
Start with status when joining a project, resuming work, or deciding what needs attention.
It assesses the available records and evidence, explains what remains, and recommends a next
action. It also works when the project has no Lusona records yet. Asking for status does
not authorize new implementation work.
For a new idea, explore helps examine possibilities and tradeoffs; frame-product helps
establish the users, problem, and product direction. As that direction becomes clearer,
map-capabilities identifies what the product needs to support and useful delivery order.
specify clarifies behavior and acceptance criteria; architect resolves consequential
technical decisions about boundaries, contracts, and ownership. design focuses on UI
mockups, screen flows, and diagrams; you can request it directly without another stage first.
implement carries an authorized change through execution and verification.
These are common ways work develops, not required stages. Start with the group that fits the
current need, skip settled questions, and return to any group when evidence calls for it.
Investigation can precede a specification or interrupt implementation; review can reveal a
reason to revisit a design.
| Action group | Skills and responsibilities |
| Orient | status: assess progress, remaining work, and the next useful action |
| Develop direction | explore: examine possibilities and tradeoffs frame-product: establish users, outcomes, and boundaries map-capabilities: map coverage, dependencies, and delivery order |
| Define a change | specify: agree on behavior, acceptance criteria, and TDD policy architect: resolve consequential technical decisions about boundaries, contracts, and ownership |
Lusona organizes coding-agent skills around a shared work loop:
This is a common interaction loop. Start where needed and revisit steps as evidence changes.
Choose, act, check, repeat. Choose the next action from the outcome and current
evidence. Revisit exploration, investigation, design, implementation, or review as needed.
There is no fixed skill sequence.
Revise decisions as you learn. When evidence challenges an earlier decision, revisit
it within the agreed authority. Update affected records so subsequent work follows the
revised decision.
Continue within authorization. Resolve routine choices and keep working without
repeated approval. Ask when a decision exceeds existing authority.
Verify completion. Complete the agreed scope, required checks and review, and any
required acceptance or delivery. Report what remains when completion is blocked.
For Claude Code, Codex, Cursor, Xcode 27 coding agents, and Google Antigravity App, CLI, and IDE · Checked-out source version 2.0.1
> [!WARNING]
> Do not install Lusona using Vercel Skills (npx skills@latest); use the native plugin installation.
Read why we built Lusona Skills: Why we built Lusona Skills.
This README describes the source in this checkout. GitHub installation uses committed,
pushed sources; local changes do not establish a published release.
## Index
Where to start
Choose a command
Installation
How the work continues
Questions and evidence
Recovery and resumption
Results and handoffs
Reference
Contributing
License
Development
## Where to start
Start with status when joining a project, resuming work, or deciding what needs attention.
It assesses the available records and evidence, explains what remains, and recommends a next
action. It also works when the project has no Lusona records yet. Asking for status does
not authorize new implementation work.
For a new idea, explore helps examine possibilities and tradeoffs; frame-product helps
establish the users, problem, and product direction. As that direction becomes clearer,
map-capabilities identifies what the product needs to support and useful delivery order.
specify clarifies behavior and acceptance criteria; architect resolves consequential
technical decisions about boundaries, contracts, and ownership. design focuses on UI
mockups, screen flows, and diagrams; you can request it directly without another stage first.
implement carries an authorized change through execution and verification.
These are common ways work develops, not required stages. Start with the group that fits the
current need, skip settled questions, and return to any group when evidence calls for it.
Investigation can precede a specification or interrupt implementation; review can reveal a
reason to revisit a design.
| Action group | Skills and responsibilities |
| Orient | status: assess progress, remaining work, and the next useful action |
| Develop direction | explore: examine possibilities and tradeoffs frame-product: establish users, outcomes, and boundaries map-capabilities: map coverage, dependencies, and delivery order |
| Define a change | specify: agree on behavior, acceptance criteria, and TDD policy architect: resolve consequential technical decisions about boundaries, contracts, and ownership |
收录徽章（可选）
想把收录标记放进 README 就复制这行，可选，不影响收录。
[](https://anclum.com/zh/s/lusona-ai-skills/)
- 来源
- github · https://github.com/Lusona/ai-skills
- 按领域浏览
- Agent 技能
- Traction
- 3 ★
- 收录于
- 今天
同类推荐
所属领域与标签
每个标签都能点进去看同类的全部条目。数字是当前库里的真实条数。