Agent Smith Plugins

Plugins for the Agent Smith terminal client. This repository contains a collection of feature extensions organized into four categories: AI Agents, Code Management, System Utilities, and Web Capabilities.

Documentation for Humans

• Full Documentation: https://lynxai-team.github.io/agent-smith/plugins — Complete plugin documentation for users
• Core Framework: Agent Smith — Runtime, CLI, and core packages

Documentation for AI Agents

• AGENTS.md: entrypoint for navigation
• .agents/documentation/codebase-summary.md: codebase navigation

Plugins Overview

🤖 AI Agents

Plugin	Version	Description	Documentation
agents		Multi-agent coordination system with 16 specialized agents (coordinator, search, code, doc, sql, help variants), 4 workflows, and 15+ skills for task management. Provides context helper fragments and collaborative agent orchestration.	doc

💻 Code Management

Plugin	Version	Description	Documentation
git		Git operations with AI-powered commit message generation. Supports diff analysis, commit workflows, and package-specific commits through 9 specialized agents.	doc
sqlite		SQLite database operations including schema extraction, read-only queries, and confirmed read/write execution. Features workflow orchestration with AI agent support.	doc

⚙️ System Utilities

Plugin	Version	Description	Documentation
fs		Filesystem operations (read, write, list directories) with path authorization security checks. Provides read-only and read/write agent configurations for safe file interaction.	doc
shell		Sandboxed command execution in Docker containers using SimpleBox (general shell) and CodeBox (Python). Includes security evaluation, complexity routing, and script generation tasks.	doc

🌐 Web Capabilities

Plugin	Version	Description	Documentation
search		Multi-backend web search with support for DuckDuckGo, smolagents, crawl4ai (with JS rendering), and Wikipedia. Includes browser automation via Playwright and inference augmentation agents.	doc
video		YouTube video transcript extraction and AI chat interaction over video content. Supports configurable model parameters for transcript-based conversations.	doc

---

Architecture Highlights

• Plugin-based extensibility: Each plugin is an independent package that registers commands, actions, agents, or tasks with the Agent Smith CLI framework via YAML definitions.

• Workflow orchestration: Plugins use YAML-defined workflows to chain actions (shell/DB/search operations) with AI agents (LLM-based generation), enabling multi-step pipelines.

• Security-first design:

  • System plugins enforce path authorization (fs plugin)
  • Sandboxed execution via Docker containers (shell plugin)
  • Read-only modes and user confirmation prompts for write operations (sqlite plugin)

• Multi-backend support: Search plugin provides redundant search backends (DuckDuckGo, smolagents, crawl4ai, Wikipedia) with agent-driven orchestration.

Dependencies

Plugins have varying dependencies based on their functionality:

• git: Uses @agent-smith/cli, commander, @inquirer/prompts, @inquirer/select
• sqlite: Uses @agent-smith/cli, better-sqlite3, @inquirer/prompts, @inquirer/select
• fs: Uses Node.js built-in modules (fs, path)
• shell: Uses @boxlite-ai/boxlite for Docker containerized execution
• search: Uses ddgs, smolagents, crawl4ai, playwright, youtube_transcript_api
• video: Uses youtube_transcript_api
• agents: No external dependencies — agents, workflows, and skills are loaded by the core framework

Related Repositories

• @agent-smith/core — Core framework providing workflow engine, agent runtime, and tool integration
• @agent-smith/cli — Terminal client that consumes these plugins