A
abdurshd/skills
Portable Agent Skills for orchestration, review, design, video production, internationalization, and shipping
在 GitHub 查看 ↗链接已于 2026-11-14 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": {
"abdurshd-skills": {
"command": "npx",
"args": ["skills", "add", "abdurshd/skills"]
}
}
}解析自我们于 2026-11-13 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": {
"abdurshd-skills": {
"command": "npx",
"args": ["skills", "add", "abdurshd/skills"]
}
}
}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 abdurshd-skills -- npx skills add abdurshd/skills
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
- 2 ★ · abdurshd/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 2026-11-14 · 最近见到 2026-11-14 · 共见到 1 次 · 被 1 个来源列出
- 评分构成
- 综合分 103 · 2 ★ · 1 个来源 · repo
- 安装
npx skills add abdurshd/skills
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 add abdurshd/skills。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
abdurshd/skills；我们收录于 2026-11-14；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。# Portable agent skills for coding workflows
以下摘录自该仓库的 README（片段） · 在 GitHub 读完整 README →
Reusable Agent Skills for implementation, review, design, video production, startup listings, internationalization, and shipping. Each skill is a self-contained folder with a standard SKILL.md plus optional scripts and references.
The workflow belongs to the skill, not to the application hosting it. The same folder can run in Claude Code, Codex, Cursor, or another compatible coding agent when that client exposes the capabilities the workflow needs.
## Portability contract
Every portable skill follows these rules:
1. Explicit user selection wins. A tool, model, coding agent, or host named by the user overrides the defaults.
2. Project/session configuration comes next. Reuse an already configured reviewer or worker when the user did not choose one.
3. Defaults fill only unspecified roles. Names such as codex-review, fable-review, and fable-opus describe the default pairing, not a mandatory host.
4. No silent substitution. If an explicitly requested agent is unavailable, report the missing executable, authentication, model access, or harness capability.
5. Capabilities, not product tool names. Skills ask for files, shell, browser, isolated workers, or read-only review. Each host maps those capabilities to its native tools.
Host, role, provider, and model are separate choices. For example, Cursor can host fable-opus while Opus performs implementation, or Claude Code can host codex-review while Codex CLI performs the independent review.
## Current coding and reasoning models
Verified against the official provider catalogs on September 26, 2026; Anthropic rows re-verified November 14, 2026. This list covers the general-purpose models relevant to these workflows; speech, image, and other specialized models are selected separately.
| Provider | Model | API model ID | Role |
| OpenAI | GPT-6 Astra | gpt-6-astra | Flagship reasoning, orchestration, and independent review |
| OpenAI | GPT-6 Sol | gpt-6-sol | Coding and agent workflows balancing capability and cost |
| OpenAI | GPT-6 Luna | gpt-6-luna | Focused tasks and high-volume work |
| Anthropic | Claude Fable 5.2 | claude-fable-5-2 | Demanding reasoning and long-running orchestration |
| Anthropic | Claude Opus 5.5 | claude-opus-5-5 | Implementation and independent review |
| Anthropic | Claude Sonnet 5.5 | claude-sonnet-5-5 | Everyday coding and agent work |
| Anthropic | Claude Haiku 5.5 | claude-haiku-5-5 | Simple tasks where speed and cost matter |
Sources: OpenAI model catalog, Anthropic model catalog, and Claude Code model configuration. Claude Mythos 5.2 is available only to verified organizations and is not a default for these skills.
Claude Code requires v2.1.280 or later for Opus 5.5; older versions may resolve fable to Fable 5.1 instead of Fable 5.2. Alias resolution varies by provider, gateway, environment overrides, and CLI version. Confirm the resolved model; when an alias selects an older release, use the current full model ID or the provider's deployment ID. User-selected models and effort levels continue to override defaults.
## Skills
Choose a skill based on what you need next. These 20 skills cover planning, implementation, independent review, HTML review, design, video creation, startup and connector listings, and delivery.
The workflow belongs to the skill, not to the application hosting it. The same folder can run in Claude Code, Codex, Cursor, or another compatible coding agent when that client exposes the capabilities the workflow needs.
## Portability contract
Every portable skill follows these rules:
1. Explicit user selection wins. A tool, model, coding agent, or host named by the user overrides the defaults.
2. Project/session configuration comes next. Reuse an already configured reviewer or worker when the user did not choose one.
3. Defaults fill only unspecified roles. Names such as codex-review, fable-review, and fable-opus describe the default pairing, not a mandatory host.
4. No silent substitution. If an explicitly requested agent is unavailable, report the missing executable, authentication, model access, or harness capability.
5. Capabilities, not product tool names. Skills ask for files, shell, browser, isolated workers, or read-only review. Each host maps those capabilities to its native tools.
Host, role, provider, and model are separate choices. For example, Cursor can host fable-opus while Opus performs implementation, or Claude Code can host codex-review while Codex CLI performs the independent review.
## Current coding and reasoning models
Verified against the official provider catalogs on September 26, 2026; Anthropic rows re-verified November 14, 2026. This list covers the general-purpose models relevant to these workflows; speech, image, and other specialized models are selected separately.
| Provider | Model | API model ID | Role |
| OpenAI | GPT-6 Astra | gpt-6-astra | Flagship reasoning, orchestration, and independent review |
| OpenAI | GPT-6 Sol | gpt-6-sol | Coding and agent workflows balancing capability and cost |
| OpenAI | GPT-6 Luna | gpt-6-luna | Focused tasks and high-volume work |
| Anthropic | Claude Fable 5.2 | claude-fable-5-2 | Demanding reasoning and long-running orchestration |
| Anthropic | Claude Opus 5.5 | claude-opus-5-5 | Implementation and independent review |
| Anthropic | Claude Sonnet 5.5 | claude-sonnet-5-5 | Everyday coding and agent work |
| Anthropic | Claude Haiku 5.5 | claude-haiku-5-5 | Simple tasks where speed and cost matter |
Sources: OpenAI model catalog, Anthropic model catalog, and Claude Code model configuration. Claude Mythos 5.2 is available only to verified organizations and is not a default for these skills.
Claude Code requires v2.1.280 or later for Opus 5.5; older versions may resolve fable to Fable 5.1 instead of Fable 5.2. Alias resolution varies by provider, gateway, environment overrides, and CLI version. Confirm the resolved model; when an alias selects an older release, use the current full model ID or the provider's deployment ID. User-selected models and effort levels continue to override defaults.
## Skills
Choose a skill based on what you need next. These 20 skills cover planning, implementation, independent review, HTML review, design, video creation, startup and connector listings, and delivery.
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/abdurshd-skills/)
- 来源
- github · https://github.com/abdurshd/skills
- 按领域浏览
- Agent 技能
- Traction
- 2 ★
- 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。