Skip to content

Navigation Menu

Sign in
Appearance settings

Search code, repositories, users, issues, pull requests...

Provide feedback

We read every piece of feedback, and take your input very seriously.

Saved searches

Use saved searches to filter your results more quickly

Sign up
Appearance settings

Surgical integration tool - extract features from agent plugins without bloat

License

Notifications You must be signed in to change notification settings

shihwesley/agent-reverse

Repository files navigation

AgentReverse

npm MCP License: MIT

Extract tools, skills, and plugins from any source — GitHub repos, local configs, articles — and install only what you need. No cloning entire repositories for one function.

> /agent-reverse analyze https://github.com/anthropics/claude-code-plugins
Scanning repo... found 12 capabilities:
 SKILLS (5)
 code-review Review PRs for bugs, security, performance [/command]
 test-runner Run and report test suites [/command]
 feature-dev Guided feature development [/command]
 explanatory-style Educational output mode [agent-only]
 learning-style Quiz-based learning mode [agent-only]
 MCP SERVERS (2)
 typescript-lsp TypeScript language intelligence
 pyright-lsp Python type checking
 Security scan: all clear (0 critical, 0 medium)
 Which capabilities would you like to install?

Table of Contents

Features

  • Analyze anything — GitHub repos, local files, Cursor configs, web articles, your own Claude Code setup
  • Security scanning — 6-category pattern matching on every install (remote exec, data exfil, secret access, persistence, destructive ops, obfuscation)
  • Manifest trackingagent-reverse.json pins every installed capability to a source and commit, like package-lock.json for skills
  • Cross-agent restore — back up from Claude Code, restore into Cursor or Antigravity (and vice versa)
  • Subagent delegation — heavy operations (repo cloning, AST parsing, manifest diffing) run in an isolated agent-reverse-engine subagent
  • Local introspectionanalyze local inventories your environment, finds dead skills, deprecated settings, unreachable MCP servers
  • Changelog awareness — detects Claude Code version bumps, auto-applies safe migrations, flags breaking changes

Install

As a Claude Code plugin (recommended)

/plugin marketplace add shihwesley/shihwesley-plugins
/plugin install agent-reverse@shihwesley-plugins

This installs the MCP server, skills, agents, and hooks automatically.

Via npx (standalone)

# Add MCP server
claude mcp add agent-reverse -- npx -y @shihwesley/agent-reverse agent-reverse-server
# Install skills and agents
npx -y @shihwesley/agent-reverse setup

Restart Claude Code after setup.

Global install or from source

Global install:

npm install -g @shihwesley/agent-reverse
claude mcp add agent-reverse -- agent-reverse-server
agent-reverse setup

From source:

git clone https://github.com/shihwesley/agent-reverse.git
cd agent-reverse
npm install && npm run build
claude mcp add agent-reverse -- node dist/server.js
node dist/cli.js setup

Requirements: Node.js >= 20.0.0

Quick Start

Analyze a repo and install one capability:

# 1. Point at any source
/agent-reverse analyze https://github.com/cool-user/search-plugin
# 2. Install what you need
/agent-reverse install search-lite
# 3. Check your setup
/agent-reverse audit

Health-check your local environment:

/agent-reverse analyze local

Returns: installed skills count, dead skills, deprecated settings, MCP server health, and optimization suggestions.

Commands

Command What it does
/agent-reverse analyze <source> Analyze a repo, file, local env, or article URL for extractable capabilities
/agent-reverse install <id> Install a capability with security scanning
/agent-reverse sync Reinstall all manifest capabilities (new machine setup)
/agent-reverse audit Find bloat, duplicates, dead skills, deprecated configs
/agent-reverse check-updates Check for capability updates and Claude Code version changes
/agent-reverse backup [--gist] Export setup to file, Gist, or GitHub repo
/agent-reverse restore <source> Restore from backup, optionally to a different agent platform

Source types for analyze: GitHub URL, local file path, local (your environment), article URL.

Security Scanning

Every install runs 6 checks before writing files:

Category Severity Action
Remote execution CRITICAL Blocked
Data exfiltration CRITICAL Blocked
Secret access MEDIUM Warn, ask to confirm
Persistence MEDIUM Warn, ask to confirm
Destructive operations LOW Reported
Obfuscation LOW Reported

Critical findings block the install. Use --force to override after reviewing the code.

Agent Portability

Agent Config path Status
Claude Code .claude/skills/ Supported
Cursor .cursor/rules/ Supported
Antigravity skills/ Supported

Backup from one, restore to another:

/agent-reverse backup --gist
/agent-reverse restore <gist-url> --target cursor

Backup and Restore

# Local backup
/agent-reverse backup
# GitHub Gist (private)
/agent-reverse backup --gist
# Restore with merge
/agent-reverse restore backup.json --merge
# Cross-agent dry run
/agent-reverse restore backup.json --target cursor --dry-run

What gets backed up: manifest, skills, commands, rules, CLAUDE.md, and caches.

Flag Description
--gist Upload to private GitHub Gist
--gist --gist-public Upload to public Gist
--repo owner/name Push to a GitHub repo
--target <agent> Convert to: claude-code, cursor, antigravity
--merge Merge with existing setup instead of replacing
--dry-run Preview without writing

MCP Tools Reference

30 tools organized by function:

Family Tools Purpose
repo_* fetch, analyze, cleanup Clone and analyze GitHub repos
manifest_* add, remove, get, list, audit, sync, check_updates Manage capability manifest
install_capability Install with security gate
conflict_* check, resolve Detect and resolve skill conflicts
deps_* check, resolve Dependency analysis
observer_* log, get_patterns, clear Workflow pattern observation
suggester_* check, add_repo, list_repos Capability suggestions
web_* fetch, interpret Extract skills from articles
backup_* create, restore, list Backup and restore
security_scan Pattern-based security analysis
local_* scan, optimize Environment introspection
changelog_check Version-aware auto-migration

Documentation

  • PRD — vision and roadmap
  • Tech spec — architecture and JSON schemas
  • Test cases — real-world usage scenarios

Contributing

PRs for new agent adapters or language parsers welcome. See the tech spec for architecture details.

git clone https://github.com/shihwesley/agent-reverse.git
cd agent-reverse
npm install && npm run build
npm test

License

MIT

AltStyle によって変換されたページ (->オリジナル) /