AgentSkillsVault

Personal vault of Agent skills sourced from local filesystem (not the hub). Skills are grouped by category; each category folder contains one directory per skill with SKILL.md plus optional references/, templates/, assets/, scripts/, examples/ subdirs.

Layout

AgentSkillsVault/
├── README.md
├── KNOWLEDGE.md
├── agents/                          ← agent autonomy & persistence patterns
│   └── goal-persistence/            ← Codex product contract (row + idle + anti-drift)
├── devops/                          ← orchestration, kanban, cron pipelines
│   ├── kanban-orchestrator/
│   ├── kanban-cron-overseer/
│   └── cron-pipeline-state-machine/
├── software-development/            ← code review, self-review, research analysis
│   ├── aisides-ai-self-review/
│   ├── debugger-persona/            ← debug diagnosis + Chinese doc writing + code discipline persona
│   ├── software-dev-loop/           ← goal → docs → test → memory dev loop
│   ├── technical-research-analysis-doc/  ← TorchLight 调研分析文档格式 (背景→结论先行→影响链条→mermaid→方案对比)
│   ├── skill-auto-improve-mirror/   ← improve + version + mirror skills to cross-harness ~/.agents/skills/
│   ├── governance-doc-design/       ← author governance/rule/design docs (buckets + 状态 + 事实来源 + decision log)
│   └── journey/                     ← two-column ME/YOU project history + vibe-coding lessons
├── FDE/                             ← Forward Deployed Engineer / personal knowledge workflows
│   ├── book-chapter-to-vault/       ← .docx book chapter → learning notes + cheatsheet + kanban + mindmap
│   └── analysis-to-vault/           ← long-form analytical article → thesis + argument map + cheatsheet + day-job notes
├── game-dev/                        ← game design + 3D/web game tooling
    ├── game-dev-loop/               ← orchestrator: classify → GDD/TDD → scaffold → implement → test/debug → quality → guide → memory
    ├── cat-game-architecture/
    ├── guide-from-probes/           ← deterministic E2E probes → playthrough guide (no screenshots needed)
    ├── kimi3-game-gen/
    ├── gdd-markdown-template/
    ├── intro-scene-until-perfect/
    ├── phaser-gamedev/
    ├── playwright-testing/
    ├── render-quality-loop/         ← screenshot sweep → render-quality evaluation → improve loop
    ├── single-file-html-game/
    ├── technical-design-document/
    ├── three-pbr-workflow/
    └── webapp-testing/
├── math/                            ← rigorous math proof generation + adversarial audit
│   └── rigorous-proof/
├── management/                      ← 职场向上汇报 (ManageUp) + 管理方法论 (mgmt-skill) + 角色顾问
    ├── management-loop/             ← router: classify → report/advice/knowledge/methodology → apply → verify → memory
    ├── manage-up-core/ + 9 汇报技能 (weekly-report, project-update, style-report 等)
    ├── mgmt-discipline/ mgmt-individual/ mgmt-org/  (方法论知识库)
    └── bezos-advisor/ god-leader-advisor/ renzhengfei-advisor/ zhangyiming-advisor/  (角色顾问)
└── content/                         ← 博客/长文 → 社媒内容再利用 (Content Repurposing)
    ├── blog-to-twitter-post/
    ├── blog-to-linkedin-post/
    ├── long-blog-to-viral-social-posts/
    ├── tech-design-to-zhihu/        ← 技术设计稿 → 知乎专栏（无 mermaid，出 PNG）
    ├── gracker-writing/             ← 技术文章写作（准确/有用/易读，Android/性能优化长文）
    ├── shuorenhua/                  ← 去 AI 味/说人话（先保信息，再谈风格）
    ├── notes-on-writing/            ← Michael Nielsen 写作笔记（目的/标题/开头/危险词/结构/经典风格）
    ├── social-push/                 ← 一句话发布内容到多平台（小红书/X/知乎/微博/微信/掘金/Linux.do）
    ├── content-craft/               ← 内容生产总纲（提炼自全部内容 skill 的共同手艺）
    └── agent-browser/               ← agent-browser CLI 浏览器自动化（social-push 依赖）
└── design/                          ← anti-slop 前端设计 taste（Leonxlnx/taste-skill 精选 6/13）
    ├── taste-frontend/              ← 旗舰 v2：读 brief → 推断设计语言 → 出不像模板的界面
    ├── taste-redesign/              ← 现有项目：先审计 UI → 再修布局/间距/层级/样式
    ├── taste-output/                ← 反截断：强制完整输出，禁占位注释
    ├── taste-imagegen-web/          ← 网页 comps 图（每 section 一张横向图）
    ├── taste-imagegen-mobile/       ← 移动端 screen/flow 图
    └── taste-brandkit/              ← 品牌 kit 板（logo/配色/字体/身份应用）

Skills

Skill	Category	Used By
agents/goal-persistence	agents	persistent goal: durable state + idle self-start + anti-drift steering + resume
devops/kanban-orchestrator	devops	AgentMOD caretaker, AISides caretaker, PKB caretaker
devops/kanban-cron-overseer	devops	AgentMOD overseer
devops/cron-pipeline-state-machine	devops	(reference pattern)
software-development/aisides-ai-self-review	software-development	AISides caretaker
software-development/debugger-persona	software-development	debug diagnosis + Chinese doc writing + code discipline persona
software-development/software-dev-loop	software-development	goal → plan doc → implement → test → journey → memory dev loop (holy grail)
software-development/technical-research-analysis-doc	software-development	UE pak/打包/热更/渲染调研 → TorchLight 格式分析文档（背景→结论先行→影响链条→mermaid→方案对比）
software-development/journey	software-development	two-column ME/YOU project history + vibe-coding lessons
software-development/skill-auto-improve-mirror	software-development	improve-in-place + version + mirror skills to cross-harness ~/.agents/skills/ (Kilo Code, Cursor, Claude Code)
software-development/llm-friendly-dsl-verification	software-development	prove an LLM-friendly DSL end-to-end (compile→run→semantic-golden→fix-feedback→scene) with real evidence
software-development/governance-doc-design	software-development	author governance/rule/design docs: product/design/dev/legal buckets, 状态+读取时机+事实来源 skeleton, append-only decision log
FDE/book-chapter-to-vault	FDE	.docx book chapter → learning notes + cheatsheet + kanban + mindmap
FDE/analysis-to-vault	FDE	long-form analytical article → thesis + argument map + day-job notes
game-dev/cat-game-architecture	game-dev	C.A.T / GDC 2026 AI-driven 3D game refactor
game-dev/game-dev-loop	game-dev	orchestrator: classify → GDD/TDD → scaffold → implement → test/debug → quality → guide → memory
game-dev/guide-from-probes	game-dev	deterministic E2E probes (assertions = ground truth) → human-readable playthrough guide
game-dev/kimi3-game-gen	game-dev	KIMI3 vibecoding-webapp-swarm (parallel coder agents)
game-dev/gdd-markdown-template	game-dev	13-section GDD scaffold
game-dev/intro-scene-until-perfect	game-dev	scope-cut to ONE scene + infinite polish loop (GDD + Art Book + Code Book)
game-dev/phaser-gamedev	game-dev	Phaser 3 2D games: scenes, sprites, physics, tilemaps, animations
game-dev/playwright-testing	game-dev	Playwright MCP / Vitest / Jest frontend + canvas game testing
game-dev/render-quality-loop	game-dev	screenshot sweep → honest render-quality evaluation → improve code loop
game-dev/single-file-html-game	game-dev	mini-browser-games style single-file HTML games + tier audit
game-dev/technical-design-document	game-dev	TDD scaffold (engineering counterpart to GDD)
game-dev/three-pbr-workflow	game-dev	token-friendly Three.js PBR scene scaffolding
game-dev/webapp-testing	game-dev	Playwright verification of running web game
math/rigorous-proof	math	rigorous math proof generation + adversarial audit (theorem → complete proof → compilable LaTeX)
management/manage-up-core	management	ManageUp 核心方法论（反空话五大原则：BLUF/数据锚定/So-What/行动导向/校准语言）
management/management-loop	management	router: classify → report/advice/knowledge/methodology → apply (anti-fluff) → verify → memory
management/weekly-report	management	数据驱动周报/月报
management/project-update	management	项目进展汇报（红绿灯+风险矩阵）
management/performance-review	management	绩效自评/述职
management/proposal	management	提案/资源申请（ROI + 不行动代价）
management/meeting-summary	management	会议纪要（聚焦决策与行动项）
management/quarterly-review	management	季度复盘/QBR（记分卡+根因分析）
management/upward-email	management	向上汇报邮件
management/one-on-one-prep	management	1:1 沟通准备
management/style-report	management	大厂汇报风格（阿里/亚马逊/字节/谷歌/微软/腾讯，合并自 6 个 style-*）
management/mgmt-discipline	management	15 管理学科横切方法论知识库
management/mgmt-individual	management	管理大师/实践者个人方法论（德鲁克/波特/科特勒等）
management/mgmt-org	management	17 标杆企业组织方法论（华为/字节/亚马逊/丰田等）
management/bezos-advisor / god-leader-advisor / renzhengfei-advisor / zhangyiming-advisor	management	角色化管理顾问
content/blog-to-twitter-post	content	博客 → Twitter/X 帖子（源锚定 + 趋势 + 平台规则 + 视觉简报）
content/blog-to-linkedin-post	content	博客 → LinkedIn 思想领导力帖子（3 版本 + 推荐）
content/long-blog-to-viral-social-posts	content	长文 → 病毒式社媒帖子（钩子模板 + 平台规则）
content/tech-design-to-zhihu	content	技术设计稿 → 知乎专栏：说人话、多 PNG、TL;DR + 收获结论 + PS/PPS/PPPS
content/gracker-writing	content	技术文章写作：准确/有用/易读，翻译腔修正 + 四层质检（Android/性能优化/Perfetto）
content/shuorenhua	content	去 AI 味/说人话：中文 AI 腔清理，保真合同 + Tier/档位/scope + 回读质检
content/notes-on-writing	content	Michael Nielsen 写作笔记：目的/标题/开头/危险词/结构/经典风格（英文）
content/social-push	content	一句话发布内容到多平台（小红书/X/知乎/微博/微信/掘金/Linux.do），基于 agent-browser
content/agent-browser	content	agent-browser CLI 浏览器自动化（快照/ref/登录态，social-push 依赖）
content/content-craft	content	内容生产总纲：准确/有用/易读 + 活人感 + 去 AI 味 + 四形态骨架 + 发布红线（提炼自全部内容 skill）
design/taste-frontend	design	anti-slop 前端（旗舰 v2）：读 brief → 推断设计语言 → VARIANCE/MOTION/DENSITY 三旋钮 → 出不像模板的界面
design/taste-redesign	design	现有项目升级：审计 UI → 识别 generic AI 模式 → 高级设计标准（不破坏功能）
design/taste-output	design	反截断：完整代码输出、禁占位注释、token 拆分处理
design/taste-imagegen-web	design	网页 comps 参考图（每 section 单独一张横向图，供编码模型复刻）
design/taste-imagegen-mobile	design	移动端 screen/flow 参考图（iOS/Android/跨平台）
design/taste-brandkit	design	品牌 kit 板：logo 方向/配色/字体/身份应用

评分（0 / 10 / 20）

逐份通读全部 skill 后打分：0 无用 · 10 专业 · 20 天才。结果：16 个 20 分（天才）、81 个 10 分（专业）、0 个 0 分——重复/破损已在前几轮清理掉，剩下没有「无用」的。

20 分 · 天才（最值得用）

Skill	类别	为什么是天才
`goal-persistence`	agents	持久目标的非显然洞察：durable state + idle self-start + anti-drift
`rigorous-proof`	math	对抗式审计的数学证明法
`shuorenhua`	content	去 AI 味的 Tier/scope/protected-span 系统
`manage-up-core`	management	反空话替换表 + 先问再写，精炼方法论
`god-leader-advisor`	management	带出处纪律的元顾问 + 生命线模型
`workflows`	qiushi	跨 skill 编排 + 显式数据交接
`mass-line`	qiushi	「群众路线」→ AI 信息源映射，真正新颖
`journey`	software-development	ME/YOU 双列 + vibe-coding 论纲
`grilling`	MattSkills	frontier 访谈轮次 + 事实/决策分离
`diagnosing-bugs`	MattSkills	先反馈环后假设的调试法
`wayfinder`	MattSkills	战争迷雾决策图，plan-don't-do
`cat-game-architecture`	game-dev	GDC 2026 C.A.T 提炼成可 grep 验证的规则
`guide-from-probes`	game-dev	断言而非像素才是 ground truth
`intro-scene-until-perfect`	game-dev	单场景打磨的 scope 纪律
`game-template`	opengame-harness	物理态家族演化 + 稳定性指标
`game-template-engineering`	opengame-harness	KEEP/COPY hook 纪律 + 反幻觉规则

10 分 · 专业（81 个）

其余全部为 10 分——清晰、可执行、做好本职，是扎实的默认项。不在上面 16 个里的，都是 10 分。

See KNOWLEDGE.md for operational findings: Claude CLI adoption status, parallelism config, worker profile setup, and live board state.

Installation

Every skill is a complete directory containing SKILL.md and, when present, its references/, scripts/, templates/, assets/, and examples/. Always copy the whole skill directory. Copying only SKILL.md can leave broken links and missing helper scripts.

The examples below install one skill. Replace <category> and <skill-name> with a vault path such as game-dev/phaser-gamedev.

Quick reference

CLI / agent	Project-level destination	User/global destination	Refresh / verify
Claude Code	`<project>/.claude/skills/<skill-name>/`	`~/.claude/skills/<skill-name>/`	Start a new session or reload Claude Code
Codex CLI	`<project>/.codex/skills/<skill-name>/`	`~/.codex/skills/<skill-name>/`	Start a new Codex session
Kilo Code CLI	`<project>/.kilo/skills/<skill-name>/`	`~/.kilo/skills/<skill-name>/`	Run `/reload` or start a new session
Hermes Agent	N/A; use an external directory for shared project skills	`~/.hermes/skills/<category>/<skill-name>/`	`hermes skills list`
Portable Agent Skills	`<project>/.agents/skills/<skill-name>/`	`~/.agents/skills/<skill-name>/`	Restart the compatible agent host

Windows: ~ means your user profile directory, normally C:\Users\<you>. Git Bash examples can use /d/GitRepo-AI/AgentSkillsVault; PowerShell examples should use D:\GitRepo-AI\AgentSkillsVault.

Claude Code

Project installation (recommended for repository-specific skills):

$Vault = "D:\GitRepo-AI\AgentSkillsVault"
$Project = "C:\path\to\project"
$Skill = "game-dev\phaser-gamedev"
$Name = Split-Path $Skill -Leaf
New-Item -ItemType Directory -Force "$Project\.claude\skills" | Out-Null
Copy-Item -Recurse -Force "$Vault\$Skill" "$Project\.claude\skills\$Name"

Global installation for all projects:

$Vault = "D:\GitRepo-AI\AgentSkillsVault"
$Skill = "game-dev\phaser-gamedev"
$Name = Split-Path $Skill -Leaf
New-Item -ItemType Directory -Force "$HOME\.claude\skills" | Out-Null
Copy-Item -Recurse -Force "$Vault\$Skill" "$HOME\.claude\skills\$Name"

Reload Claude Code or begin a new session after installation. A project may intentionally ignore .claude/; check git status if the skill should be shared with the team.

Codex CLI

Project installation:

$Vault = "D:\GitRepo-AI\AgentSkillsVault"
$Project = "C:\path\to\project"
$Skill = "game-dev\phaser-gamedev"
$Name = Split-Path $Skill -Leaf
New-Item -ItemType Directory -Force "$Project\.codex\skills" | Out-Null
Copy-Item -Recurse -Force "$Vault\$Skill" "$Project\.codex\skills\$Name"

Global installation:

$Vault = "D:\GitRepo-AI\AgentSkillsVault"
$Skill = "game-dev\phaser-gamedev"
$Name = Split-Path $Skill -Leaf
New-Item -ItemType Directory -Force "$HOME\.codex\skills" | Out-Null
Copy-Item -Recurse -Force "$Vault\$Skill" "$HOME\.codex\skills\$Name"

Start a new Codex session so it rescans the skill directories.

Kilo Code CLI

Install the CLI if needed:

npm install -g @kilocode/cli

Project installation:

$Vault = "D:\GitRepo-AI\AgentSkillsVault"
$Project = "C:\path\to\project"
$Skill = "game-dev\phaser-gamedev"
$Name = Split-Path $Skill -Leaf
New-Item -ItemType Directory -Force "$Project\.kilo\skills" | Out-Null
Copy-Item -Recurse -Force "$Vault\$Skill" "$Project\.kilo\skills\$Name"

Global installation:

$Vault = "D:\GitRepo-AI\AgentSkillsVault"
$Skill = "game-dev\phaser-gamedev"
$Name = Split-Path $Skill -Leaf
New-Item -ItemType Directory -Force "$HOME\.kilo\skills" | Out-Null
Copy-Item -Recurse -Force "$Vault\$Skill" "$HOME\.kilo\skills\$Name"

Kilo also discovers .agents/skills/ by default and can read .claude/skills/ when Claude Code compatibility is enabled. Run /reload in Kilo after copying a skill. For a shared vault without copying files, add an absolute path under skills.paths in kilo.jsonc:

{
  "skills": {
    "paths": ["D:/GitRepo-AI/AgentSkillsVault/game-dev"]
  }
}

Hermes Agent

Hermes uses ~/.hermes/skills/ as its primary skill store and supports category directories:

$Vault = "D:\GitRepo-AI\AgentSkillsVault"
$Category = "game-dev"
$Name = "phaser-gamedev"
New-Item -ItemType Directory -Force "$HOME\.hermes\skills\$Category" | Out-Null
Copy-Item -Recurse -Force "$Vault\$Category\$Name" "$HOME\.hermes\skills\$Category\$Name"

Verify discovery:

hermes skills list

To let Hermes read the vault directly instead of maintaining copies, add the category directories you use to ~/.hermes/config.yaml:

skills:
  external_dirs:
    - D:/GitRepo-AI/AgentSkillsVault/game-dev
    - D:/GitRepo-AI/AgentSkillsVault/devops
    - D:/GitRepo-AI/AgentSkillsVault/software-development
    - D:/GitRepo-AI/AgentSkillsVault/FDE

Hermes can modify skills in writable external directories. Keep the vault in Git and review changes before committing. If a skill is published through a supported Hermes Hub or URL source, prefer hermes skills inspect followed by hermes skills install; manually copied vault skills can still be listed and loaded normally.

Portable `.agents/skills` installation

For hosts supporting the open Agent Skills compatibility directory, install project skills under .agents/skills/:

$Vault = "D:\GitRepo-AI\AgentSkillsVault"
$Project = "C:\path\to\project"
$Skill = "game-dev\phaser-gamedev"
$Name = Split-Path $Skill -Leaf
New-Item -ItemType Directory -Force "$Project\.agents\skills" | Out-Null
Copy-Item -Recurse -Force "$Vault\$Skill" "$Project\.agents\skills\$Name"

This is useful when one repository is used with multiple compatible agents. Support and precedence vary by host, so use a host-native directory when exact behavior matters.

Install an entire category

The following PowerShell helper copies every immediate skill directory that contains a SKILL.md:

$Vault = "D:\GitRepo-AI\AgentSkillsVault"
$Category = "game-dev"
$Destination = "$HOME\.claude\skills" # Change for Codex, Kilo, etc.
New-Item -ItemType Directory -Force $Destination | Out-Null
Get-ChildItem "$Vault\$Category" -Directory | Where-Object {
  Test-Path (Join-Path $_.FullName "SKILL.md")
} | ForEach-Object {
  Copy-Item -Recurse -Force $_.FullName (Join-Path $Destination $_.Name)
}

Dependencies, MCP servers, and plugins

Skills provide instructions; they do not automatically install the tools they describe. Read each skill's SKILL.md, README.md, or adjacent installation guide and install only the required dependencies. For example, the Phaser skills have a dedicated installation guide covering Playwright MCP, Pillow, and Phaser.

Review third-party scripts and MCP servers before enabling them. Do not copy credentials, private assets, generated caches, or machine-specific secrets into the vault.

Sync Back to Vault

When installed skills are updated, sync back to the correct category folder:

# From installed location to vault
cp -r ~/AppData/Local/hermes/skills/<category>/<skill-name>/ \
  /d/GitRepo-My/AgentSkillsVault/<category>/<skill-name>/

# Verify byte-identical
diff -rq <vault-copy> <installed-copy>

Navigation Menu

Repository files navigation

AgentSkillsVault

Layout

Skills

评分（0 / 10 / 20）

20 分 · 天才（最值得用）

10 分 · 专业（81 个）

Installation

Quick reference

Claude Code

Codex CLI

Kilo Code CLI

Hermes Agent

Portable `.agents/skills` installation

Install an entire category

Dependencies, MCP servers, and plugins

Sync Back to Vault

About

Releases

Packages

Contributors

Languages

Latest commit

History

Folders and files

Repository files navigation

AgentSkillsVault

Layout

Skills

评分（0 / 10 / 20）

20 分 · 天才（最值得用）

10 分 · 专业（81 个）

Installation

Quick reference

Claude Code

Codex CLI

Kilo Code CLI

Hermes Agent

Portable .agents/skills installation

Install an entire category

Dependencies, MCP servers, and plugins

Sync Back to Vault

About

Resources

Stars

Watchers

Forks

Releases

Packages

Contributors

Languages

Portable `.agents/skills` installation