Code Pluginsource linked

SmartContext — Intelligent Conversation Guidelinesv1.0.7

SmartContext — Intelligent Conversation Guidelines / 智能对话准则

@smartcontext/openclaw-plugin·runtime smartcontext·by @hherosoul
Community code plugin. Review compatibility and verification before install.
openclaw plugins install clawhub:@smartcontext/openclaw-plugin
Latest release: v1.0.7Download zip

Capabilities

configSchema
Yes
Executes code
Yes
HTTP routes
0
Runtime ID
smartcontext

Compatibility

Built With Open Claw Version
2026.4.2
Min Gateway Version
2026.4.2
Plugin Api Range
>=2026.4.2
Plugin Sdk Version
2026.4.2
Security Scan
VirusTotalVirusTotal
Benign
View report →
OpenClawOpenClaw
Benign
high confidence
Purpose & Capability
Name/description match implementation: the plugin registers a before_prompt_build hook, loads domain SKILL.md files, composes tiered guidelines, and exposes commands to list/add/remove skills and role tags. It requests no unrelated credentials or binaries and only requires filesystem access to the plugin state and skills directories, which is appropriate for its purpose.
Instruction Scope
SKILL.md files are passed essentially raw into the LLM prompt (the README and code explicitly state and implement this). The install command copies arbitrary SKILL.md content from a provided filesystem path into the plugin's skills directory, after only a minimal validation (non-empty). This is coherent with the plugin goal but means installing an untrusted SKILL.md can inject arbitrary system-level instructions into every conversation — a useful feature for customization but a potential avenue for prompt-injection if you install untrusted skills.
Install Mechanism
There is no external download or remote install happening in the plugin bundle. The package contains source and compiled files; runtime behavior relies on the OpenClaw plugin API. No install spec pulls code from untrusted URLs. The only filesystem operations are local reads/writes (skills directory, state config), which matches the plugin's functionality.
Credentials
The plugin declares no required environment variables or external credentials. It reads and writes a config JSON in the runtime state directory and copies SKILL.md files into its skills directory; this filesystem access is proportional to the stated capability. There are no requests for unrelated secrets or network credentials.
Persistence & Privilege
The plugin is not enabled by default and does not set always:true. It registers a before_prompt_build hook (expected for a guideline injector) and stores installed skills and config in its state directory. It does not modify other plugins' configuration. Autonomous invocation via the hook is the designed behavior and is consistent with its purpose.
Assessment
This skill appears to do what it says: inject tiered, domain-specific guidelines into prompts. However, its install command copies arbitrary SKILL.md files into the plugin's skills directory and those files are sent nearly verbatim into the model context. Only install SKILL.md files from trusted sources and review their content before installing — a malicious or poorly written SKILL.md can change model behavior across all conversations. Also be aware the plugin stores installed skills and its configuration in the gateway's state directory; avoid placing sensitive secrets inside SKILL.md files. If you need stricter control, keep the plugin disabled by default and enable it only when you trust the active skills.

Verification

Tier
source linked
Scope
artifact only
Summary
Validated package structure and linked the release to source metadata.
Commit
plugins
Tag
plugins
Provenance
No
Scan status
clean

Tags

latest
1.0.7

SmartContext — Intelligent Conversation Guidelines / 智能对话准则

Professional Conversation Navigation System: Injects domain-specific, tiered importance guidelines into every conversation, helping LLM focus on critical information and improve the quality and efficiency of professional dialogues. / 专业对话导航系统:向每一次对话注入领域特定的、分级的重要性准则,帮助 LLM 专注于关键信息,提升专业对话的质量和效率。


Table of Contents / 目录


Overview / 项目概述

SmartContext is a professional conversation navigation system designed for OpenClaw. Through a Plugin + Skill hybrid architecture, it enables flexible extension and dynamic injection of domain-specific rules.

SmartContext 是为 OpenClaw 设计的专业对话导航系统,通过 Plugin + Skill 的混合架构,实现领域特定规则的灵活扩展和动态注入。

Core Value / 核心价值

  • Precision Navigation / 精准导航: Based on 5-tier importance system (Tier 0-4), helps LLM understand what matters / 基于 5 层重要性分级体系 (Tier 0-4),帮助 LLM 理解什么重要
  • Domain Adaptation / 领域适配: Pluggable Skill architecture, supports any professional domain / 可插拔的 Skill 架构,支持任意专业领域
  • Manual Control / 手动控制: User explicitly specifies active domains, precise control over rule application / 用户显式指定活跃领域,精确控制规则生效
  • Non-intrusive Integration / 无侵入集成: Injects guidelines via before_prompt_build hook, does not affect memory-core / 通过 before_prompt_build 钩子注入准则,不影响 memory-core
  • Multi-domain Collaboration / 多领域协作: Supports cross-domain collaboration, automatically handles rule conflicts / 支持跨领域协作,自动处理规则冲突

Key Features / 核心特性

1. 5-Tier Priority System / 5 层优先级分级体系

Tier / 层级Description / 描述Example / 示例
Tier 0Always Prioritize (Never Overlook) / 始终优先(绝不能忽视)User personas, architectural decisions, strong constraints / 用户画像、架构决策、强约束
Tier 1Highest Priority Focus / 最高优先级关注Code review, performance optimization, security design / 代码审查、性能优化、安全设计
Tier 2High Priority Attention / 高优先级注意Requirements doc, detailed design, test cases / 需求文档、详细设计、测试用例
Tier 3Keep In Mind / 记住这些Coding conventions, project config, debug records / 编码规范、项目配置、调试记录
Tier 4Can Reference Briefly / 可以简要参考General knowledge, confirmatory responses, duplicate content / 通用知识、确认回复、重复内容

2. Plugin + Skill Hybrid Architecture / Plugin + Skill 混合架构

SmartContext Plugin (Framework / 框架)
├── Core: Guideline Injection Engine / 核心:准则注入引擎
├── Config: Domain Selection Management / 配置:领域选择管理
└── Extension: Skill Loader / 扩展:Skill 加载器

Skill (Domain Rules / 领域规则)
├── General Purpose / 通用
├── Software Engineering / 软件工程
├── Bioinformatics / 生物信息学
└── Floriculture / 花艺

3. Multi-domain Composition / 多领域组合

Supports activating multiple domains simultaneously, automatically handles rule conflicts and cross-domain collaboration.

支持同时激活多个领域,自动处理规则冲突和跨领域协作。


Quick Start / 快速开始

Installation / 安装

Install the plugin using OpenClaw's plugin manager:

使用 OpenClaw 的插件管理器安装:

openclaw plugins install clawhub:smartcontext/smartcontext

Or using the package name directly:

或者直接使用包名:

openclaw plugins install @smartcontext/openclaw-plugin

After installation, restart the OpenClaw gateway:

安装完成后,重启 OpenClaw gateway:

openclaw gateway restart

Basic Usage / 基本使用

# View current configuration / 查看当前配置
/smartcontext-config

# List available skills / 列出可用领域
/smartcontext-list-skills

# Switch to specified skill (replaces others) / 切换到指定领域(替换其他)
/smartcontext-use software-engineering

# Add skill / 添加领域
/smartcontext-add bioinformatics

# Remove skill / 移除领域
/smartcontext-remove software-engineering

# Set role tags (overwrites) / 设置角色标签(覆盖)
/smartcontext-set-role frontend typescript react

# Add role tags / 添加角色标签
/smartcontext-add-role backend

# Remove role tags / 移除角色标签
/smartcontext-remove-role frontend

# Clear all role tags / 清除所有角色标签
/smartcontext-clear-role

# Pin important content / 标记重要内容
/smartcontext-pin "This project must support offline functionality / 本项目必须支持离线功能"

# Show pinned content / 显示固定内容
/smartcontext-show-pinned

# Unpin at index (1-based) / 取消固定指定索引(从 1 开始)
/smartcontext-unpin-at 1

# Unpin all / 取消所有固定
/smartcontext-unpin-all

# Reset to default config / 重置为默认配置
/smartcontext-reset

# Install new skill / 安装新技能
/smartcontext-install-skill my-domain /path/to/SKILL.md

# Uninstall skill / 卸载技能
/smartcontext-uninstall-skill my-domain

Usage Guide / 使用指南

Configuration / 配置说明

Config Item / 配置项Type / 类型Description / 说明Default / 默认值
activeDomainsstring[]List of active domain IDs / 启用的领域 ID 列表[]
roleTagsstring[]User role tags / 用户角色标签[]
pinnedItemsstring[]User-pinned important content / 用户手动标记的重要内容[]

Built-in Skills / 内置领域

1. General Purpose / 通用

Suitable for all scenarios, provides basic conversation guidelines.

适用于所有场景,提供基础的对话准则。

Key focus / 核心重点:

  • User's core needs and goals / 用户的核心需求和目标
  • Core constraints that must be followed / 必须遵守的核心约束
  • Critical decisions and important choices / 关键决策和重要选择

2. Software Engineering / 软件工程

Covers the complete lifecycle including requirements analysis, architecture design, coding implementation, testing validation, deployment, and operations.

覆盖完整的软件开发生命周期,包括需求分析、架构设计、编码实现、测试验证、部署运维。

Key focus / 核心重点:

  • User personas, core usage scenarios, key business rules / 用户画像、核心使用场景、关键业务规则
  • Confirmed architectural decisions: technology choices, architectural style, system boundaries, module division / 已确认的架构决策:技术选型、架构风格、系统边界、模块划分
  • Code review comments and fix records / 代码审查意见和修复记录
  • Security design and privacy protection measures / 安全设计和隐私保护措施
  • Database schemas and API contracts / 数据库 schema 和 API 契约

3. Bioinformatics / 生物信息学

Integrates biology, computer science, mathematics, and statistics to analyze and interpret biological data.

整合生物学、计算机科学、数学和统计学,用于分析和解释生物数据。

Key focus / 核心重点:

  • Research objectives, core biological hypotheses, key experimental designs / 研究目标、核心生物学假设、关键实验设计
  • Confirmed analysis pipelines: tool chains, parameter configurations, quality control standards / 已确认的分析流程:工具链、参数配置、质量控制标准
  • Genomic sequence alignment results and variant calling records / 基因组序列比对结果和变异检测记录
  • Patient consent information and data privacy protocols / 患者知情同意信息和数据隐私协议

4. Floriculture / 花艺

Integrates horticulture, botany, and environmental science to optimize plant growth, health, and aesthetics.

整合园艺学、植物学和环境科学,优化植物生长、健康和美观。

Key focus / 核心重点:

  • Gardening objectives, core plant health goals, key cultivation strategies / 园艺目标、核心植物健康目标、关键栽培策略
  • Confirmed care regimes: species-specific requirements, critical growth parameters, pest management protocols / 已确认的养护方案:物种特定要求、关键生长参数、病虫害管理协议
  • Plant identification results and species-specific care requirements / 植物鉴定结果和物种特定养护要求
  • Pest and disease diagnosis records and treatment protocols / 病虫害诊断记录和治疗方案

Skill Extension / 技能扩展

Creating a New Skill / 创建新技能

Important / 重要:

  • The skill file must be named SKILL.md (exact name, case-sensitive) / 技能文件必须命名为 SKILL.md(精确名称,大小写敏感)
  • You can write any format that makes sense to LLM / 你可以写任何 LLM 能理解的格式
  • The plugin directly passes the raw content to LLM / 插件直接把原始内容传给大模型

Quick Steps / 快速步骤

# 1. Write your SKILL.md file (any format works!) / 编写你的 SKILL.md 文件(任何格式都可以!)

# 2. Install the skill / 安装技能
/smartcontext-install-skill my-domain /path/to/your/SKILL.md

# 3. Enable the skill / 启用技能
/smartcontext-add my-domain

SKILL.md Template / SKILL.md 模板

You can use this as a starting point, but feel free to modify it in any way that works for your domain:

你可以以此为起点,但可以根据你的领域自由修改:

# SmartContext Skill: Your Domain Name / 你的领域名称

## Domain Summary / 领域摘要

Brief description of what this domain is about and what it covers.

简要说明这个领域是关于什么的,涵盖哪些内容。

## Priority Declaration / 优先级声明

When rules from this domain conflict with other domains, priority: High/Medium/Low

当此领域的规则与其他领域冲突时,优先级:高/中/低

## Priority Guidelines (Tier 0 → Tier 4) / 优先级指南 (Tier 0 → Tier 4)

### Tier 0 — Always Prioritize (Never Overlook) / 始终优先(绝不能忽视)
- Most important content / 最重要的内容
- Things that must never be overlooked / 绝不能忽视的事情

### Tier 1 — Highest Priority Focus / 最高优先级关注
- Very important content / 非常重要的内容
- Things that should be focused on / 应该重点关注的事情

### Tier 2 — High Priority Attention / 高优先级注意
- Important content / 重要内容
- Things that need attention / 需要注意的事情

### Tier 3 — Keep In Mind / 记住这些
- Useful information / 有用的信息
- Things to remember / 要记住的事情

### Tier 4 — Can Reference Briefly / 可以简要参考
- Less important content / 不太重要的内容
- Things that can be referenced briefly / 可以简要参考的事情

## Dynamic Adjustment Rules / 动态调整规则

- Any dynamic rules you want / 任何你想要的动态规则

## Domain-Specific Rules / 领域特定规则

- Any domain-specific rules / 任何领域特定的规则

## Cross-Domain Collaboration / 跨领域协作

- How to collaborate with other domains / 如何与其他领域协作

Reference Existing Skills / 参考现有技能

For real-world examples, check the built-in skills in the skills/ directory:

想要真实示例,请查看 skills/ 目录下的内置技能:

  • skills/smartcontext-general-purpose/SKILL.md
  • skills/smartcontext-software-engineering/SKILL.md
  • skills/smartcontext-bioinformatics/SKILL.md
  • skills/smartcontext-floriculture/SKILL.md

Uninstalling a Skill / 卸载技能

# Uninstall a skill / 卸载技能
/smartcontext-uninstall-skill my-domain

Project Structure / 项目结构

smartcontext-plugin/
├── openclaw.plugin.json          # Plugin Manifest / 插件清单
├── package.json                  # npm package config / npm 包配置
├── tsconfig.json                 # TypeScript config / TypeScript 配置
├── README.md                     # This file / 本文件
│
├── src/
│   ├── index.ts                  # Plugin entry / 插件入口
│   ├── types.ts                  # Global type definitions / 全局类型定义
│   │
│   ├── core/                     # Core modules / 核心模块
│   │   ├── guideline-engine.ts   # Guideline generation engine / 准则生成引擎
│   │   ├── skill-loader.ts       # Skill discovery & loading / Skill 发现与加载
│   │   ├── domain-composer.ts    # Multi-domain rule composition / 多领域规则组合
│   │   ├── prompt-builder.ts     # Prompt text builder / Prompt 文本构建器
│   │   ├── skill-discovery.ts    # Skill discoverer / Skill 发现器
│   │   └── skill-manager.ts      # Skill manager / Skill 管理器
│   │
│   ├── commands/                 # User commands / 用户命令
│   │   └── index.ts              # Command registration / 命令注册
│   │
│   ├── config/                   # Config management / 配置管理
│   │   ├── schema.ts             # Config schema / 配置 Schema
│   │   ├── defaults.ts           # Default config values / 默认配置值
│   │   └── store.ts              # Config storage / 配置存储
│   │
│   └── utils/                    # Utilities / 工具
│       ├── markdown-parser.ts    # Simple Skill markdown parser / 简单的 Skill Markdown 解析器
│       ├── logger.ts             # Logger / 日志工具
│       └── cache.ts              # Cache / 内存缓存
│
├── skills/                       # Built-in skills / 内置技能
│   ├── smartcontext-general-purpose/
│   ├── smartcontext-software-engineering/
│   ├── smartcontext-bioinformatics/
│   └── smartcontext-floriculture/
│
└── dist/                         # Build output / 构建输出

Development / 开发指南

Local Development / 本地开发

# 1. Install dependencies / 安装依赖
npm install

# 2. Build the project / 构建项目
npm run build

Build for Release / 发布构建

# Build the project / 构建项目
npm run build

Boundaries with OpenClaw / 与 OpenClaw 的职责边界

Responsibility / 职责OpenClawSmartContext Plugin
Context Management / 上下文管理✅ Full responsibility / 完全负责❌ Not involved / 不介入
Memory Compression / 内存压缩✅ Full responsibility / 完全负责❌ Not involved / 不介入
Prompt Building / Prompt 构建✅ Full responsibility / 完全负责✅ Inject system context / 注入系统上下文
Domain Rule Extension / 领域规则扩展❌ Not provided / 不提供✅ Plugin + Skill architecture
User Config Management / 用户配置管理❌ Not provided / 不提供✅ Command-line interface / 命令行接口

License / 许可证

This project is licensed under the MIT License.

本项目采用 MIT 许可证。


SmartContext — Make professional conversations more efficient! / 让专业对话更高效!