agentsclimarketplace
Sign in

Opencode user prohiler

Skill jnzhanglujia/opencode-user-prohiler

Stores and retrieves user preferences, tech stack, constraints, and project history to enable personalized AI planning, architecture design, and code generation.From its SKILL.md

Install
npx -y skills add jnzhanglujia/opencode-user-prohiler

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its file declares

Copied from the file, not written here

The file declares its own license as MIT. That is the author’s claim about this one file, and it is not the same thing as the license GitHub reports for the repository, which is listed with the other numbers below.

SKILL.md

9.4 KB, ~2.9k tokens by cl100k_base, as published. Nobody here has run it

User Context Profile Skill

🎯 Purpose

This skill provides a single source of truth for the user's personal background, technical preferences, workflow habits, constraints, and historical decisions. When asked to plan, scaffold, or build software, always consult this profile first before generating code, architecture, or tooling recommendations.

⚡ Auto-Load Requirement

此 Skill 必须在每次对话开始时自动加载。 配置方式：

  1. 确保 user-context-profile 已放入全局 skill 目录（~/.opencode/skills/）
  2. 在项目根目录的 AGENTS.md 中添加指令： 
    每次对话开始时，自动执行以下步骤：
1. 加载 `user-context-profile` skill
2. 读取 `user-data.json` 中的用户画像数据
3. 根据画像为用户提供个性化服务
  3. 加载后立即读取 user-data.json，如果已有完整画像则直接进入 Mode B（Normal Usage），无需再次提问

📋 Two Modes

Mode A: First-Time Setup (Profile Survey)

When this skill detects user-data.json has empty/missing personal fields (e.g., personal_info.gender is empty), enter Survey Mode and administer the full questionnaire below.

Mode B: Normal Usage (Profile Loaded)

When user-data.json has complete profile data, skip the survey and go directly to Load Context:

  1. Load Context: Read user-data.json at the start of any planning/building session.
  2. Extract Relevant Fields: Match the request against personal background + tech_stack + preferences + constraints + projects.
  3. Generate Output: Tailor architecture, dependencies, UI/UX patterns, code style, and dev workflows to the stored profile.
  4. Learn & Update: If the user explicitly states a new preference or corrects a choice, update user-data.json immediately.
  5. Never Assume: Do not invent preferences. If a field is missing or "unknown", ask the user or use safe defaults.

📝 Survey Mode: Questionnaire

禁止使用 question 工具。全程自然对话，每次只问1个问题，用户回答后立即记录。全部问完后统一提交保存。

分组提问顺序

第一个问题必须先问称呼。 用户回答后立即记录到 personal_info.name，后续所有问题都用此称呼来叫用户。

组1: 基本信息

  1. 称呼/名字: 开放填写（提示：你希望我怎么称呼你？可以是真名、昵称、英文名都可以）
  2. 性别: 选项→ 男 / 女 / 其他 / 不愿透露
  3. 年龄范围: 选项→ 18岁以下 / 18-25岁 / 26-35岁 / 36-45岁 / 46-55岁 / 56-65岁 / 65岁以上
  4. 身高: 开放填写（提示：单位cm，如 175）
  5. 体重: 开放填写（提示：单位kg，如 70）
  6. 最高学历: 选项→ 高中及以下 / 大专 / 本科 / 硕士 / 博士 / 其他

组2: 职业与工作

  1. 职业/岗位: 开放填写（提示：告诉我你的行业和具体岗位，比如「互联网-后端开发」「教育-教师」「自由职业-设计」）
  2. 开发角色: 选项→ 前端开发 / 后端开发 / 全栈开发 / DevOps / 数据/AI / 移动端 / 非开发角色 / 其他
  3. 工作方式偏好: 选项→ 独立开发 / 团队协作 / 管理统筹 / 混合
  4. 每日可投入工作时长: 开放填写（提示：单位小时，如 6）
  5. 精力最佳时段: 选项→ 早晨 / 下午 / 晚上 / 深夜 / 不固定

组3: 技术与工具

  1. 主力操作系统: 选项→ Windows / macOS / Linux / 多系统混用
  2. 主力编辑器/IDE: 选项→ VS Code / Vim/Neovim / JetBrains系列 / Sublime / 其他
  3. 前端框架偏好: 选项→ React / Vue / Angular / Svelte / 不用前端框架 / 不确定
  4. 后端语言偏好: 选项→ Python / JavaScript/TypeScript / Go / Rust / Java / C# / PHP / 其他 / 不确定
  5. 数据库偏好: 选项→ PostgreSQL / MySQL / MongoDB / SQLite / Redis / 其他 / 不确定
  6. 技术整体水平自评: 选项→ 新手 / 初中级 / 高级 / 专家
  7. 最佳学习方式: 选项→ 阅读文档/书籍 / 看视频教程 / 动手做项目 / 跟人讨论 / 混合

组4: 个性与偏好

  1. 性格倾向: 选项→ 内向 / 中性 / 外向
  2. 沟通风格偏好: 选项→ 直接简明 / 详细全面 / 可视化图表 / 要点列表
  3. 兴趣爱好: 开放填写（提示：告诉我你的兴趣爱好，用逗号分隔，例如「编程、游戏、跑步、摄影、读书」。越详细越好，这有助于我了解你的整体风格）

组5: 项目与约束

  1. 通常做什么类型项目: 开放填写（提示：描述你常做的项目类型，例如「Web全栈应用」「数据爬虫」「开源工具库」「AI demo」。每个类型用逗号分隔）
  2. 预算/成本倾向: 选项→ 纯开源免费 / 有小额预算 / 企业级预算 / 不限
  3. 对项目性能的期望: 开放填写（提示：描述你的性能要求，例如「首屏<2s」「API<100ms」「没特别要求」「高峰期需支持1万QPS」。越具体越好）
  4. 是否需要无障碍支持: 选项→ 必须（WCAG AA） / 尽量做到 / 暂不需要

提交保存流程

所有25个问题问完后，按以下步骤执行：

  1. 展示摘要: 列出用户所有回答的完整汇总
  2. 请求确认: "以上是你的完整画像，请确认无误。输入「提交」保存，或告诉我需要修改哪一项"
  3. 用户说「提交」后:
    • 将所有数据写入 user-data.json
    • 写入后立即读取验证
    • 输出确认信息：✅ 用户画像已保存至 user-data.json
  4. 用户要求修改某项: 回到对应问题重新确认 → 更新记录 → 再次展示摘要 → 重复第2步

关键规则：

  • 任何时候用户说「跳过」「这一项不填」→ 该字段设为 "unknown"，继续下一题
  • 任何时候用户说「退出」「不做了」→ 停止问卷，已记录的数据丢弃，不保存
  • 用户提交后必须验证 JSON 已正确写入，确认无误才算完成

📂 Data Structure (user-data.json)

{
  "schema_version": "2.0",
  "personal_info": {
    "name": "鲁佳",
    "gender": "男",
    "age_range": "26-35岁",
    "height_cm": 175,
    "weight_kg": 70,
    "education": "本科",
    "hobbies": ["编程", "游戏", "跑步"],
    "personality": "中性",
    "communication_style": "直接简明",
    "learning_style": "动手做项目",
    "best_work_hours": "6",
    "best_work_time": "晚上",
    "occupation": "互联网-后端开发",
    "dev_role": "全栈开发",
    "work_style": "团队协作"
  },
  "tech_stack": {
    "os": "macOS",
    "ide": "VS Code",
    "frontend": ["React", "TypeScript", "TailwindCSS", "Vite"],
    "backend": ["FastAPI", "PostgreSQL", "Redis"],
    "frontend_framework_pref": "React",
    "backend_language_pref": "Python",
    "database_pref": "PostgreSQL",
    "infra": ["Docker", "GitHub Actions", "Vercel"],
    "testing": ["pytest", "Playwright"],
    "avoid": ["Angular", "Java", "Webpack"],
    "package_manager": "pnpm",
    "skill_level": "高级"
  },
  "preferences": {
    "ui_style": "minimalist, dark mode default",
    "architecture": "modular, feature-based folders, prefer functional components",
    "error_handling": "fail-fast with user-friendly fallbacks",
    "documentation": "inline comments + README.md + OpenAPI spec",
    "git_workflow": "conventional commits, squash merge"
  },
  "constraints": {
    "budget": "open_source_only",
    "performance_target": "< 2s initial load, < 100ms API latency",
    "compliance": "WCAG 2.1 AA",
    "accessibility": "必须（WCAG AA）"
  },
  "project_types": ["Web全栈应用", "开源工具库"],
  "projects": [],
  "history": []
}

🔧 How to Use the Profile

When user asks you to build/plan something:

  1. Read profile from user-data.json
  2. Cross-reference user's request with their profile:
    • Use their preferred tech stack
    • Follow their communication style
    • Respect their constraints (budget, performance, accessibility)
    • Reference their past projects if relevant
  3. Explain your decisions by referencing their profile: "根据你的偏好，我用了 React + FastAPI..."
  4. If user contradicts profile, ask if they want to update: "你之前偏好 X，这次要改成 Y 吗？我来更新配置"

🔧 Cross-Agent Compatibility

  • Uses only standard JSON
  • Zero external dependencies
  • Follows Agent Skills Open Standard
  • Works in sandboxed or full-filesystem environments

🔐 Privacy & Safety

  • All data stored locally only
  • Never transmit raw profile to external APIs
  • User may edit user-data.json manually at any time
  • Sensitive fields (e.g., API keys) should be stored in .env, not here

📥 Installation

AgentPath
OpenClaw~/.openclaw/skills/user-context-profile/
OpenCode~/.config/opencode/skills/user-context-profile/
Hermes~/.hermes/skills/user-context-profile/
Claude Code~/.claude/skills/user-context-profile/
Codex CLI$CODEX_HOME/skills/user-context-profile/

What ships with it: 5 files

9.2 KB alongside SKILL.md, 1 of them executable

scripts/

Keep looking

Skills are one crate of 326,367. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.