Usage & Billing
Aurirlk-DSH-skills
Aurirlk/Aurirlk-DSH-skills · ★ 0
Aurirlk 验证可用的 DSH(DeepSeek Harness)技能合集:每个技能一个目录,装上本包即注册全部技能。当前收录 disk-butler(Windows 电脑管家:查清空间去向、按安全等级清理、Junction 透明迁移缓存、留档成本机档案)。
Manifest details
安装
dsh plugin --profile web add 'github:Aurirlk/Aurirlk-DSH-skills#c345c61b92aac2de1673c47d086fe8568f0643af'GitHub 源码安装:首次需按提示配置 allowBuilds 构建授权后重试
For install and environment setup, plus plugin development tutorials, see DSH community docs ↗
README
Table of contents
- Skills
- System tools
- Documents & diagrams
- Front-end (Nuxt 4 / Vue 3 / Vuetify 3)
- Job hunting / interviews
- Related projects
- Installation
- Option 1 — install the whole collection (recommended)
- Option 2 — link a single skill
- Option 3 — other agents
- Verifying the install
- Repository layout
- Why skills live under skills/<name>/ instead of the repository root
- Adding a skill
- Contributing
- Security
- License
My DSH skill-sharing repository. One directory per skill, each with bilingual READMEs covering usage and installation.
Two kinds of content live here:
- Original — written and used by me (e.g. disk-butler)
- Adapted — built from someone else's or an official published skill, reworked to the DSH spec
Every adapted skill records its source, licence and exactly what we changed in
ATTRIBUTIONS.md.
English | 中文
Skills
System tools
| Skill | What it does | Platform | Docs |
|---|---|---|---|
| disk-butler | Windows computer butler: find out where C/D/E space went → clean junk by safety tier → migrate hard-coded caches transparently with NTFS junctions → keep a durable machine archive | Windows | English · 中文 |
Documents & diagrams
| Skill | What it does | Origin | Docs |
|---|---|---|---|
| drawio | Generate / export / lint draw.io diagrams: automatic layered layout from a JSON spec, page-by-page export to PNG/SVG/PDF, and pre-export checks for overlapping boxes and edges routed through boxes | original | English · 中文 |
Front-end (Nuxt 4 / Vue 3 / Vuetify 3)
| Skill | What it does | Origin | Docs |
|---|---|---|---|
| code-review-refactor | Actionable code review and refactoring: duplicated logic, CSS de-duplication, config-driven pages | adapted | English · 中文 |
| git-commit-message-generator | Conventional Commits messages with subjects that actually say something | adapted | English · 中文 |
| ui-tip | One convention for global toasts: success uses defaults, errors must set the icon | adapted (body changed) | English · 中文 |
| ui-use-confirm | One convention for confirmation dialogs, plus which actions must confirm first | adapted (body changed) | English · 中文 |
| vuetify0 | Using and authoring with @vuetify/v0 headless components and composables |
official skill | English · 中文 |
Job hunting / interviews
| Skill | What it does | Origin | Docs |
|---|---|---|---|
| create-interviewer | Distils an interviewer into a continuously-evolving AI persona (tech map + behaviour archive + persona) | adapted (MIT) | English · 中文 |
| interview-skills | Big-tech AI mock interviewer: questions from company + role + JD + resume, with good-vs-bad answers and multi-round simulation | adapted (MIT) | English · 中文 |
Sources and licences are documented in
ATTRIBUTIONS.md.
Related projects
These live outside this repository (they have homes of their own — re-uploading would be pointless), so they are listed as recommendations only:
| Project | Notes |
|---|---|
| ASu-skills | A job-hunting skill collection (5.4k ⭐): resume "distillation", interview prediction and follow-up drilling, autumn recruiting tracker, open-source contribution helper. Aurirlk contributed OpenCode plugin support (PR #80) |
| @vuetify/v0 | Where this repository's vuetify0 skill comes from — the headless logic layer for Vue 3 |
Installation
This repository is a public git repository and can be used directly as a package source. No npm registry and no npm account required.
Option 1 — install the whole collection (recommended)
Add this repository as a dependency of your DSH profile to register every skill at once:
# Windows — your profile directory may not be called `desktop`; use whatever you have
cd "$env:USERPROFILE\.dsh\profiles\desktop"
pnpm add github:Aurirlk/Aurirlk-DSH-skills
# macOS / Linux
cd ~/.dsh/profiles/desktop
pnpm add github:Aurirlk/Aurirlk-DSH-skills
DSH then reads the package's dsh.bundle.patch and registers every skill under skills/.
Measured at roughly 7 seconds.
Updating to the latest version:
pnpm update dsh-aurirlk-skills
Option 2 — link a single skill
DSH discovers skills by scanning one level under a skill root (<root>/<name>/SKILL.md),
so you can link just one skill instead of the whole repository — no package install needed:
# Windows — a junction needs no administrator rights (a symlink would)
git clone https://github.com/Aurirlk/Aurirlk-DSH-skills.git D:\Dev\Aurirlk-DSH-skills
New-Item -ItemType Junction `
-Path "$env:USERPROFILE\.dsh\skills\disk-butler" `
-Target "D:\Dev\Aurirlk-DSH-skills\skills\disk-butler"
# macOS / Linux
git clone https://github.com/Aurirlk/Aurirlk-DSH-skills.git ~/dev/Aurirlk-DSH-skills
ln -s ~/dev/Aurirlk-DSH-skills/skills/disk-butler ~/.dsh/skills/disk-butler
Which option to pick:
- Option 1 installs the whole collection; skills you add later show up automatically. Good if you want to follow upstream.
- Option 2 links a single skill; the repository's other skills stay out of your environment. Good if you only want one of them.
Both use a junction/symlink, so after
git pullthe skill is already up to date — no copying. DSH watches skill roots, so added / renamed / removed skills take effect without a restart.
Option 3 — other agents
The skill body is just skills/<name>/SKILL.md and does not depend on DSH.
Any agent that understands SKILL.md (Claude Code, Codex, …) can use the directory directly.
The repository also ships .codex-plugin/plugin.json for the Codex/OpenAI plugin flow.
Verifying the install
# Just ask the agent: "my C drive is full" — it should load disk-butler automatically.
# Or run the skill's self-check by hand:
& "$env:USERPROFILE\.dsh\skills\disk-butler\scripts\Test-SkillHealth.ps1" -Deep
A note on npm: this repository's
package.jsonis a valid npm package (dsh-aurirlk-skills), but it has not been published to the npm registry, sopnpm add dsh-aurirlk-skillswill fail today. Use Option 1 instead. The git install and an npm install would ship identical contents.
Repository layout
Aurirlk-DSH-skills/
├── README.md / README.en.md ← you are here (intro + skill index)
├── LICENSE / CONTRIBUTING.md / SECURITY.md / CHANGELOG.md
├── .gitattributes / .editorconfig / .gitignore / .npmignore
├── .github/workflows/verify.yml ← CI: verifies every skill
├── scripts/Fix-Bom.ps1 ← repo-level dev tool (not published)
├── index.js / cordis.patch.yml / package.json ← collection package (registers all skills)
├── .codex-plugin/plugin.json ← Codex/OpenAI plugin manifest
└── skills/
└── disk-butler/ ← one skill
├── README.md / README.en.md ← intro / usage / installation
├── SKILL.md ← the skill body (what the agent reads)
├── scripts/ ← executable scripts
└── references/ ← docs loaded on demand
Why skills live under skills/<name>/ instead of the repository root
This is not arbitrary — it is where two specs intersect:
- DSH filesystem skill discovery scans exactly one level:
<root>/<name>/SKILL.md, no nesting - The Codex/OpenAI plugin spec requires
plugin.json'sskillsfield to point at a directory namedskills
Putting skills under skills/ satisfies both, so one source of truth ships as three packages —
a DSH plugin, a Codex plugin, and a plain agent skill — with no duplicated copies.
Adding a skill
skills/<name>/SKILL.md— the body. Frontmatter requiresname(kebab-case, must match the directory name) anddescriptionskills/<name>/README.md— the human-facing intro, usage and installation- Add a row to the Skills table above
- Run the checks:
.\scripts\Fix-Bom.ps1 # encoding / BOM
.\skills\<name>\scripts\Test-SkillHealth.ps1 -Deep # per-skill self-check (if present)
node scripts\self-test.mjs # collection registration logic
index.js needs no changes — it iterates skills/ and registers every directory containing a
SKILL.md.
Contributing
See CONTRIBUTING.md. The single most important rule is the BOM convention —
these skills target Windows PowerShell, and PS 5.1 is extremely sensitive to encoding:
.ps1 must have a BOM while SKILL.md must not, two opposite requirements that are easy to
get backwards.
Security
Skills here may modify or delete files on your machine (disk-butler cleans up disk space, for example). Each skill's own README documents its safety model — read it before use. See SECURITY.md for how to report a vulnerability.
License
MIT © Aurirlk