A
anoscranel/agent-skills-subskills
33 MIT-licensed Agent Skills for software engineering, writing and prompt engineering, plus a meta-skill that…
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
- ★ 数取值
- 0 ★ · anoscranel/agent-skills-subskills · 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 2026-10-26 · 最近见到 2026-11-14 · 共见到 23 次 · 被 3 个来源列出
- 评分构成
- 综合分 303 · 3 个来源 · repo
常见问题(答案都来自我们的记录)
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
anoscranel/agent-skills-subskills;我们收录于 2026-10-26;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。# agent-skills-subskills
以下摘录自该仓库的 README(片段) · 在 GitHub 读完整 README →
35 MIT-licensed Agent Skills for professional work: software engineering, writing, prompt engineering, and one meta-skill that builds and audits skills like these.
Each skill is a short routing index (SKILL.md) over a flat set of leaf files (subskills/.md). The agent loads the index, picks one leaf, and follows it. Every leaf holds its full procedure and never hands the task to another skill.
## What's in the pack
| Suite | Skills | Leaves |
| Software engineering | 19 | 60 |
| Writing | 14 | 48 |
| Prompt engineering (ac-prompt-engineer) | 1 | 12 |
| Skill authoring (ac-subskills-author) | 1 | 10 |
| Total | 35 | 130 |
Software engineering: agent-coding-governance · api-design · architecture · ci-and-release · code-construction · code-review · codebase-maintenance · database-changes · debugging · frontend-ui · git-workflow · performance · quality-verification · reliability · software-engineering-standards · spec-writing · task-planning · testing · design-taste
Writing: academic-writing · business-writing · drafting · editing · evidence-and-citation · narrative-writing · news-writing · persuasive-writing · plain-language · proposal-writing · technical-writing · voice-and-style · writing-standards · slide-decks
Every skill name carries the ac- prefix (for example ac-testing).
## How a leaf is built
Each leaf has a routing header (name, domain, trigger, anti-trigger, scope, status) and, for the craft leaves, six sections: Standard, Decision rules, Procedure, Anti-patterns, Done checklist, Sources. Claims cite public primary sources inline. Paywalled or single-source claims carry a ⚠️ flag.
## Install
Copy one skill directory (or all of them) into a skills directory your harness reads:
| Harness | Global path | Project path |
| opencode | ~/.config/opencode/skills/ / or ~/.agents/skills/ / | .opencode/skills/ / |
| Claude Code | ~/.claude/skills/ / | .claude/skills/ / |
| Other Agent Skills harnesses | see your harness's docs | |
sh
git clone https://github.com/anoscranel/agent-skills-subskills.git
cp -r agent-skills-subskills/skills/ac-testing ~/.claude/skills/
Restart the harness, then confirm the skill is listed (opencode: opencode debug skill). Paths per the opencode skills docs.
## Code in this repo
The only executable code is four Python checkers in skills/ac-subskills-author/scripts/; leaves may show illustrative snippets that nothing runs. The four checkers use the standard library only, make no network calls and write no files, and run only when you call them:
| Script | What it does |
| linttree.py | Checks a skill tree's structure: frontmatter, domain chain, index ↔ disk, links, no hand-off lines |
Each skill is a short routing index (SKILL.md) over a flat set of leaf files (subskills/.md). The agent loads the index, picks one leaf, and follows it. Every leaf holds its full procedure and never hands the task to another skill.
## What's in the pack
| Suite | Skills | Leaves |
| Software engineering | 19 | 60 |
| Writing | 14 | 48 |
| Prompt engineering (ac-prompt-engineer) | 1 | 12 |
| Skill authoring (ac-subskills-author) | 1 | 10 |
| Total | 35 | 130 |
Software engineering: agent-coding-governance · api-design · architecture · ci-and-release · code-construction · code-review · codebase-maintenance · database-changes · debugging · frontend-ui · git-workflow · performance · quality-verification · reliability · software-engineering-standards · spec-writing · task-planning · testing · design-taste
Writing: academic-writing · business-writing · drafting · editing · evidence-and-citation · narrative-writing · news-writing · persuasive-writing · plain-language · proposal-writing · technical-writing · voice-and-style · writing-standards · slide-decks
Every skill name carries the ac- prefix (for example ac-testing).
## How a leaf is built
Each leaf has a routing header (name, domain, trigger, anti-trigger, scope, status) and, for the craft leaves, six sections: Standard, Decision rules, Procedure, Anti-patterns, Done checklist, Sources. Claims cite public primary sources inline. Paywalled or single-source claims carry a ⚠️ flag.
## Install
Copy one skill directory (or all of them) into a skills directory your harness reads:
| Harness | Global path | Project path |
| opencode | ~/.config/opencode/skills/ / or ~/.agents/skills/ / | .opencode/skills/ / |
| Claude Code | ~/.claude/skills/ / | .claude/skills/ / |
| Other Agent Skills harnesses | see your harness's docs | |
sh
git clone https://github.com/anoscranel/agent-skills-subskills.git
cp -r agent-skills-subskills/skills/ac-testing ~/.claude/skills/
Restart the harness, then confirm the skill is listed (opencode: opencode debug skill). Paths per the opencode skills docs.
## Code in this repo
The only executable code is four Python checkers in skills/ac-subskills-author/scripts/; leaves may show illustrative snippets that nothing runs. The four checkers use the standard library only, make no network calls and write no files, and run only when you call them:
| Script | What it does |
| linttree.py | Checks a skill tree's structure: frontmatter, domain chain, index ↔ disk, links, no hand-off lines |
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/anoscranel-agent-skills-subskills/)
- 来源
- github · https://github.com/anoscranel/agent-skills-subskills
- 按领域浏览
- Agent 技能
- 收录来源
- 3 个独立来源
- 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。