███████╗██╗ ██╗██╗██╗ ██╗ ██╗ ██╗██╗ ██╗██████╗
██╔════╝██║ ██╔╝██║██║ ██║ ██║ ██║██║ ██║██╔══██╗
███████╗█████╔╝ ██║██║ ██║ ███████║██║ ██║██████╔╝
╚════██║██╔═██╗ ██║██║ ██║ ██╔══██║██║ ██║██╔══██╗
███████║██║ ██╗██║███████╗███████╗ ██║ ██║╚██████╔╝██████╔╝
╚══════╝╚═╝ ╚═╝╚═╝╚══════╝╚══════╝ ╚═╝ ╚═╝ ╚═════╝ ╚═════╝
Plug-and-play skill packs for AI Agents.
给 AI 装技能包。不调 prompt，直接上专家。
Prompt = 一句话指令 → "帮我分析这份数据" → AI 随便聊两句
Skill = 一套完整的 SOP 手册 → 流程 + 策略 + 质量红线 + 错误兜底 + 结构化输出
复制粘贴即装。 零依赖，零配置。当前适配 OpenCode，本质是 Markdown，任何支持 system prompt 的平台都能用。
git clone https://github.com/Lirsakura/skills-hub.git
cp skills-hub/agents/advisor-agent.md ~/.config/opencode/agents/
cp -r skills-hub/skills/advisor-agent/ ~/.config/opencode/skills/
# 完事。打开 OpenCode 说人话就行。
✅ 可用🚧 开发中💡 待认领 — 欢迎 PR 或 Issue 投票 👍
| 状态 | Skill | 简介 | 亮点 |
|---|---|---|---|
| ✅ | advisor-agent | 导师情报分析 | 11 源扫描 · 🔴🟡🟢 风险标记 · 对比/面试模式 |
| 💡 | paper-reader | 学术论文深度拆解 | 核心贡献 · 方法论 · 局限性 · 可复现性评估 |
| 💡 | market-analyst | 行业/产品调研 | 竞品分析 · 市场规模 · 趋势预测 |
| 状态 | Skill | 简介 | 亮点 |
|---|---|---|---|
| ✅ | code-auditor | 白盒代码安全审计 | 6 阶段流程 · 跨文件追踪 · 可执行 PoC |
| 💡 | vuln-scanner | 漏洞扫描与验证 | 自动化扫描 · CVE 匹配 · 修复建议 |
| 状态 | Skill | 简介 | 亮点 |
|---|---|---|---|
| ✅ | travel-agent | AI 旅行规划师 | 6 种模式 · 门牌号精度 · 出境三语 · MBTI 路线 |
| 💡 | meal-planner | 冰箱有什么→做什么菜 | 食材匹配 · 营养均衡 · 难度分级 |
| 💡 | fitness-coach | 个性化训练计划 | 目标导向 · 器材适配 · 渐进超负荷 |
| 状态 | Skill | 简介 | 亮点 |
|---|---|---|---|
| ✅ | env-setup | GitHub 项目一键配环境 | conda/venv/docker · GPU 自检 · 镜像回退 · 18 个子模块 |
| ✅ | git-sync | Git 仓库同步管理 | push/pull/sync-log · 冲突处理 · 日志追踪 |
| 💡 | data-analyst | 自动 EDA + 可视化 | 数据画像 · 异常检测 · 洞察报告 |
| 状态 | Skill | 简介 | 亮点 |
|---|---|---|---|
| 💡 | resume-tailor | 根据 JD 定制简历 | 逐条匹配 · ATS 友好 · 量化建议 |
| 💡 | mock-interview | 模拟技术面试 | 实时追问 · 评分反馈 · 知识点定位 |
| 状态 | Skill | 简介 | 亮点 |
|---|---|---|---|
| 💡 | news-digest | 每日多源新闻聚合 | 去重 · 摘要 · 观点对比 · 趋势追踪 |
| 💡 | apartment-hunter | 租房情报调查 | 房源分析 · 周边设施 · 差评避坑 |
🎓 advisor-agent — 选导师前必看
选错导师毁三年。 11 个平台（研控 · PI Review · 知乎 · 小红书 · RateMyProfessors · Reddit · GradCafe · PubPeer...）交叉验证，输出结构化尽职调查报告。
"查一下 MIT 的 XXX 教授" → 完整报告
"对比 A 教授和 B 教授" → 对比分析
"下周面试 XXX，帮我准备" → 面试情报
三个体量版本：完整版 49K（大模型）· 精简版 11K（中等模型）· 单文件 4K（本地小模型）
🔒 code-auditor — 人类级白盒审计
不是 grep 'eval('。6 阶段流程：代码接入 → 全量理解 → 深度分析 → 对抗验证 → 报告 → 协作。聚焦 SQLi / XSS / RCE / SSRF / 文件上传，跨文件数据流追踪，输出可执行 PoC。
✈️ travel-agent — 拿着手机就能走
6 种模式（小白探索 / 拍照穿搭 / MBTI 定制 / 出境 / 地图生成...），酒店精确到门牌号 + 差评，出境三语 + 双币种 + 签证 + 租车，预算贴合度 90%-105%。
🔧 env-setup — 给个链接就能跑
给一个 GitHub URL 或本地路径，自动识别技术栈（Python/Node/Java/Docker），搭建完整可运行环境。conda 优先，支持 GPU 自检、镜像回退、端口隔离。18 个子模块覆盖 app/db/helpers/scripts。
"帮我装一下 https://github.com/xxx/yyy" → clone + 识别 + 搭建
"当前目录的环境帮我装一下" → 就地搭建
"找个 stable diffusion webui 装一下" → 搜索 + 选择 + 搭建
🔄 git-sync — Git 操作不用记命令
push / pull / sync-log 三个子 skill，处理日常 Git 同步。自动处理冲突、生成同步日志。
开 Issue 就行。哪怕一句 "要是有个 XX 就好了" 也够。
- 标题：
[Skill Request] 你想要什么 - 👍 投票越多，越优先实现
- 想自己做？直接认领，PR 提过来
每一条 Issue 我都会看。好的 Skill 都是从一个真实痛点开始的。
💡 提想法（Issue）· 🐛 报 Bug · ⭐ Star · 📢 转发 · 📝 提交 Skill（PR）
不会写代码也能贡献——一个好需求比十个 demo 更有价值。
提交 Skill：Fork → 参考 templates/skill-template.md → PR 详见 CONTRIBUTING.md
███████╗██╗ ██╗██╗██╗ ██╗ ██╗ ██╗██╗ ██╗██████╗
██╔════╝██║ ██╔╝██║██║ ██║ ██║ ██║██║ ██║██╔══██╗
███████╗█████╔╝ ██║██║ ██║ ███████║██║ ██║██████╔╝
╚════██║██╔═██╗ ██║██║ ██║ ██╔══██║██║ ██║██╔══██╗
███████║██║ ██╗██║███████╗███████╗ ██║ ██║╚██████╔╝██████╔╝
╚══════╝╚═╝ ╚═╝╚═╝╚══════╝╚══════╝ ╚═╝ ╚═╝ ╚═════╝ ╚═════╝
Plug-and-play skill packs for AI Agents.
Prompt = one-line instruction → "analyze this data" → AI rambles
Skill = a complete SOP manual → workflow + strategy + quality gates + error handling + structured output
Copy-paste to install. Zero deps. Zero config. Built for OpenCode. Works anywhere with system prompts (it's just Markdown).
git clone https://github.com/Lirsakura/skills-hub.git
cp skills-hub/agents/advisor-agent.md ~/.config/opencode/agents/
cp -r skills-hub/skills/advisor-agent/ ~/.config/opencode/skills/
# Done. Open OpenCode. Talk normally.
✅ Available🚧 WIP💡 Open — PRs welcome or 👍 upvote to prioritize
| Status | Skill | Description | Highlights |
|---|---|---|---|
| ✅ | advisor-agent | Grad advisor intelligence | 11 sources · 🔴🟡🟢 risk flags · Compare & interview prep |
| 💡 | paper-reader | Academic paper deep-dive | Core contributions · Methodology · Limitations |
| 💡 | market-analyst | Industry/product research | Competitive analysis · Market sizing · Trends |
| Status | Skill | Description | Highlights |
|---|---|---|---|
| ✅ | code-auditor | White-box security audit | 6-stage pipeline · Cross-file tracing · Executable PoCs |
| 💡 | vuln-scanner | Vulnerability scan & verify | Auto-scan · CVE matching · Fix suggestions |
| Status | Skill | Description | Highlights |
|---|---|---|---|
| ✅ | travel-agent | AI travel planner | 6 modes · Street-level picks · Trilingual · MBTI routes |
| 💡 | meal-planner | Fridge → recipe suggestions | Ingredient matching · Nutrition · Difficulty tiers |
| 💡 | fitness-coach | Custom training plans | Goal-based · Equipment-aware · Progressive overload |
| Status | Skill | Description | Highlights |
|---|---|---|---|
| ✅ | env-setup | One-click dev env for any repo | conda/venv/docker · GPU detection · Mirror fallback · 18 submodules |
| ✅ | git-sync | Git repo sync management | push/pull/sync-log · Conflict handling · Log tracking |
| 💡 | data-analyst | Auto EDA + visualization | Data profiling · Anomaly detection · Insight reports |
| Status | Skill | Description | Highlights |
|---|---|---|---|
| 💡 | resume-tailor | Tailor resume to JD | Line-by-line match · ATS-friendly · Quantification |
| 💡 | mock-interview | Simulate tech interviews | Follow-ups · Scoring · Knowledge gap analysis |
| Status | Skill | Description | Highlights |
|---|---|---|---|
| 💡 | news-digest | Multi-source daily news | Dedup · Summaries · Opinion contrast · Trends |
| 💡 | apartment-hunter | Rental intelligence | Listing analysis · Neighborhood · Red flags |
🎓 advisor-agent — Due diligence before choosing your advisor
Bad advisor = bad grad school. Cross-references 11 platforms (PI Review · RateMyProfessors · Reddit · GradCafe · PubPeer...) into a structured due diligence report.
"Look up Professor XXX at MIT" → Full report
"Compare Professor A vs B" → Side-by-side
"Interview with XXX next week" → Interview intel
Three sizes: Full 49K (frontier) · Compact 11K (mid-range) · Single-file 4K (local models)
🔒 code-auditor — Human-level white-box audit
Not grep 'eval('. 6-stage pipeline: intake → comprehension → deep analysis → adversarial verification → report → collaboration. Targets SQLi / XSS / RCE / SSRF / file upload. Cross-file data flow tracing. Executable PoCs.
✈️ travel-agent — Follow your phone and go
6 modes (explore / photo outfits / MBTI / international / map gen...). Street-level hotel picks with negative reviews. Trilingual + dual currency + visa + car rental for international trips. Budget accuracy 90%-105%.
🔧 env-setup — Give it a URL, get a running env
Give a GitHub URL or local path, auto-detect tech stack (Python/Node/Java/Docker), build a complete runnable environment. Conda-first, GPU detection, mirror fallback, port isolation. 18 submodules covering app/db/helpers/scripts.
🔄 git-sync — Git without remembering commands
push / pull / sync-log sub-skills for everyday Git sync. Auto conflict handling and sync logging.
Open an Issue. Even "wish there was a XX" is enough.
- Title:
[Skill Request] What you want - 👍 upvotes = priority
- Want to build it yourself? Claim it and PR
I read every Issue. Every great Skill starts with a real pain point.
💡 Ideas (Issue) · 🐛 Bugs · ⭐ Star · 📢 Share · 📝 Submit a Skill (PR)
No code? No problem — a great request is worth more than ten demos.
Submit a Skill: Fork → Use templates/skill-template.md → PR See CONTRIBUTING.md
⭐ Star = 让更多人发现 · helps more people find this
💬 Open an Issue — 任何想法我都看 · I read them all
Built with obsession. Skills > Prompts.