CategoryWeightCategory scoreContribution
Securityprompt, exec, net, exfil, eval
35%
35.0 pts
Supply chainhash, typosquat, maintainer, lockfile
20%
20.0 pts
Maintenancestaleness, pinning, CI
15%
15.0 pts
TransparencySKILL.md, perms, README
15%
15.0 pts
Communityinstalls, verify, response
15%
15.0 pts
Findings & checks · 0 flagged
Security
✓—All security checks passedNo findings in this category for the latest scan.pass
Supply chain
✓—All supply chain checks passedNo findings in this category for the latest scan.pass
Maintenance
✓—All maintenance checks passedNo findings in this category for the latest scan.pass
Transparency
✓—All transparency checks passedNo findings in this category for the latest scan.pass
Community
✓—All community checks passedNo findings in this category for the latest scan.pass
Every scanned point with the score it earned and what moved between them.
9ab6e8e
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
AgentReverse
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
- Install
- Quick Start
- Commands
- Security Scanning
- Agent Portability
- Backup and Restore
- MCP Tools Reference
- Documentation
- Contributing
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 tracking —
agent-reverse.jsonpins 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-enginesubagent - Local introspection —
analyze localinventories 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-pluginsThis 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 setupRestart Claude Code after setup.
<details> <summary>Global install or from source</summary>
Global install:
npm install -g @shihwesley/agent-reverse
claude mcp add agent-reverse -- agent-reverse-server
agent-reverse setupFrom 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</details>
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 auditHealth-check your local environment:
/agent-reverse analyze localReturns: 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 cursorBackup 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-runWhat 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 testLicense
MIT