extension-agent-installer 1.5.5

Secure extension installer for AI coding agents with vulnerability scanning

extension-agent-installer

Secure extension installer for AI coding agents with vulnerability scanning and credential guidance

Current Version: 1.5.5 | Changelog | Repository

Why This Skill Exists

The AI era brings new security risks.

AI agents like Claude Code, OpenCode, and Cursor have unprecedented access to your digital life:

• Your files and documents
• Your terminal and commands
• Your API keys and credentials
• Your browsing and communication

A single malicious extension can steal your banking credentials, exfiltrate sensitive data, or execute unauthorized transactions.

This skill protects users who cannot verify extension safety themselves.

Many people use AI agents but don't have the technical expertise to:

• Review extension source code
• Identify hidden malicious payloads
• Detect prompt injection attacks
• Understand what permissions an extension requires

Our mission is to promote digital hygiene in the AI era by:

1. Never install blindly - Always scan before installation
2. Alert to risks - Show vulnerabilities and explain what they mean
3. Guide setup - Help users understand what access they're granting
4. Empower users - Give everyone the tools to make informed decisions

Don't let your AI agent become a backdoor for attackers. Verify before you install.

---

Features

• Multi-Client Support - Works with OpenCode, Claude Code, Cursor, Windsurf, Gemini CLI, Codex, GitHub Copilot
• Security Scanning - Dual vulnerability detection using mcp-scan (Snyk) and Agent Trust Hub (Gen Digital)
• Credential Guidance - Step-by-step setup help for non-technical users
• Vulnerability Notifications - Always alerts user before installing risky extensions
• Auto-Detection - Automatically determines extension type and target client
• Dependency Checking - Verifies required tools before installation
• Self-Update - Can update itself from GitHub

Updating

To update this skill to the latest version, send this to your agent:

Update extension-agent-installer from https://github.com/lmt-expert-company/extension-agent-installer

Steps:
1. Fetch SKILL.md from the repository
2. Compare version numbers
3. If newer version available, download and replace current SKILL.md
4. Also update scripts/ and references/ folders
5. Report what changed

Or simply say:

• "update extension-agent-installer"
• "check for updates"
• "обнови скилл инсталлер"

Installation

Quick Install (Recommended)

Using uv (fastest):

# Auto-detect client, install to current project
uvx extension-agent-installer install

# Install globally for all projects
uvx extension-agent-installer install --global

# Specify client
uvx extension-agent-installer install --client opencode --global

Using npm:

# Auto-detect client, install to current project
npx extension-agent-installer install

# Install globally for all projects
npx extension-agent-installer install --global

# Specify client
npx extension-agent-installer install --client opencode --global

Supported Clients

Client	--client value	Project Path	Global Path
OpenCode	opencode	.opencode/skills/	~/.config/opencode/skills/
Claude Code	claude	.claude/skills/	~/.claude/skills/
Cursor	cursor	.cursor/skills/	~/.cursor/skills/
Windsurf	windsurf	.windsurf/skills/	~/.codeium/windsurf/skills/
Gemini CLI	gemini	.gemini/skills/	~/.gemini/skills/
Codex	codex	.agents/skills/	~/.agents/skills/
GitHub Copilot	copilot	.github/skills/	~/.copilot/skills/

Detect Installed Clients

uvx extension-agent-installer detect
# or
npx extension-agent-installer detect

---

Manual Installation (Alternative)

Choose your AI client below and copy the instruction to send to your agent.

---

For OpenCode

Copy and send this to your agent:

Install the skill from repository: https://github.com/lmt-expert-company/extension-agent-installer

Steps:
1. Create directory: .opencode/skills/extension-agent-installer/ (project) or ~/.config/opencode/skills/extension-agent-installer/ (global)
2. Download SKILL.md from the repository and save it to the path above
3. Optionally download scripts/ and references/ folders for full functionality
4. Verify the file was created correctly

---

For Claude Code

Copy and send this to your agent:

Install the skill from repository: https://github.com/lmt-expert-company/extension-agent-installer

Steps:
1. Create directory: .claude/skills/extension-agent-installer/ (project) or ~/.claude/skills/extension-agent-installer/ (global)
2. Download SKILL.md from the repository and save it to the path above
3. Optionally download scripts/ and references/ folders for full functionality
4. Verify the file was created correctly

---

For Cursor

Copy and send this to your agent:

Install the skill from repository: https://github.com/lmt-expert-company/extension-agent-installer

Steps:
1. Create directory: .cursor/skills/extension-agent-installer/ (project) or ~/.cursor/skills/extension-agent-installer/ (global)
2. Download SKILL.md from the repository and save it to the path above
3. Optionally download scripts/ and references/ folders for full functionality
4. Verify the file was created correctly

---

For Windsurf

Copy and send this to your agent:

Install the skill from repository: https://github.com/lmt-expert-company/extension-agent-installer

Steps:
1. Create directory: .windsurf/skills/extension-agent-installer/ (project) or ~/.codeium/windsurf/skills/extension-agent-installer/ (global)
2. Download SKILL.md from the repository and save it to the path above
3. Optionally download scripts/ and references/ folders for full functionality
4. Verify the file was created correctly

---

For Gemini CLI

Copy and send this to your agent:

Install the skill from repository: https://github.com/lmt-expert-company/extension-agent-installer

Steps:
1. Create directory: .gemini/skills/extension-agent-installer/ (project) or ~/.gemini/skills/extension-agent-installer/ (global)
2. Download SKILL.md from the repository and save it to the path above
3. Optionally download scripts/ and references/ folders for full functionality
4. Verify the file was created correctly

---

For Codex

Copy and send this to your agent:

Install the skill from repository: https://github.com/lmt-expert-company/extension-agent-installer

Steps:
1. Create directory: .agents/skills/extension-agent-installer/ (project) or ~/.agents/skills/extension-agent-installer/ (global)
2. Download SKILL.md from the repository and save it to the path above
3. Optionally download scripts/ and references/ folders for full functionality
4. Verify the file was created correctly

---

For GitHub Copilot

Copy and send this to your agent:

Install the skill from repository: https://github.com/lmt-expert-company/extension-agent-installer

Steps:
1. Create directory: .github/skills/extension-agent-installer/ (project) or ~/.copilot/skills/extension-agent-installer/ (global)
2. Download SKILL.md from the repository and save it to the path above
3. Optionally download scripts/ and references/ folders for full functionality
4. Verify the file was created correctly

---

Usage

Once installed, just give your AI agent a link to any extension:

Install this skill: https://github.com/user/awesome-skill

The agent will:

1. Fetch and analyze the extension
2. Scan for security vulnerabilities
3. Alert you to any risks found
4. Ask for your confirmation
5. Install to the correct location
6. Guide you through credential setup if needed

Example Prompts

Install this MCP server: https://github.com/modelcontextprotocol/servers/tree/main/src/github

Add this plugin to opencode: https://github.com/user/opencode-wakatime

I want a notification plugin for my AI agent

Security

This skill uses two complementary security scanners:

1. mcp-scan (Snyk)

Local static analysis that detects:

• Prompt injection attacks
• Tool poisoning attacks
• Toxic flows
• Malware payloads
• Hard-coded secrets

GitHub Repository

2. Agent Trust Hub (Gen Digital)

Cloud-based verification backed by Gen Threat Labs:

• Real-time threat intelligence
• Community-reported vulnerabilities
• Trusted developer verification

Agent Trust Hub

Vulnerability Handling

Severity	Action
SAFE	Install freely
LOW	Warn, install
MEDIUM	Ask user
HIGH	Warn strongly, require explicit OK
CRITICAL	Strongly discourage, require explicit OK

Supported Clients

Client	Project Path	Global Path	Docs
OpenCode	.opencode/skills/	~/.config/opencode/skills/	Docs
Claude Code	.claude/skills/	~/.claude/skills/	Docs
Cursor	.cursor/skills/	~/.cursor/skills/	Docs
Windsurf	.windsurf/skills/	~/.codeium/windsurf/skills/	Docs
Gemini CLI	.gemini/skills/	~/.gemini/skills/	Docs
Codex	.agents/skills/	~/.agents/skills/	Docs
GitHub Copilot	.github/skills/	~/.copilot/skills/	Docs

Credential Setup

This skill automatically detects when extensions need credentials and guides you through setup:

1. Detects required credentials - API keys, tokens, OAuth
2. Provides direct links - URLs to get credentials
3. Step-by-step instructions - Plain language, no jargon
4. Offers to help - Can assist with configuration

Supported Services

Service	Credential Type
OpenAI	API Key
Anthropic	API Key
GitHub	Personal Access Token
Vercel	API Token
Sentry	Auth Token
Stripe	Secret Key
Cloudflare	API Token

Project Structure

extension-agent-installer/
├── SKILL.md                    # Main skill file
├── README.md                   # This file
├── LICENSE                     # MIT License
├── scripts/
│   ├── scan_extension.py       # mcp-scan wrapper
│   └── scan_agent_trust.py     # Agent Trust Hub API client
└── references/
    ├── mcp-scan.md             # mcp-scan documentation
    └── agent-trust-hub.md      # Agent Trust Hub documentation

Inspiration & Resources

This project was inspired by and builds upon:

• Snyk agent-scan - Security scanner for AI agents, MCP servers and agent skills
• Gen Digital Agent Trust Hub - AI agent security scanner
• VoltAgent awesome-agent-skills - Curated collection of 380+ agent skills
• OpenCode Documentation - Official OpenCode docs
• OpenCode Repository - Core OpenCode project and docs source
• OpenCode Skill Creator workflow (.opencode/skills/skill-creator) - primary workflow used to structure and iterate this skill during development

Contributing

Contributions are welcome. This project is developed with an agent-first workflow, and we strongly recommend following the same pattern.

Development workflow used in this project

This repository was iterated using the OpenCode Skill Creator workflow (local skill path: .opencode/skills/skill-creator) together with OpenCode docs.

If you use OpenCode while contributing:

1. Load skill-creator for skill structure and metadata validation guidance.
2. Keep installation behavior and docs aligned (README.md, SKILL.md, AGENTS.md).
3. Validate both package targets (PyPI + npm) before opening a PR.

Local validation before PR

Run these commands from repo root:

python -m build
npm pack --dry-run
python -m py_compile src/extension_agent_installer/*.py scripts/*.py
node --check src/installer.js
python src/extension_agent_installer/installer.py version
node src/installer.js version

Versioning and release discipline

• Follow semantic versioning (MAJOR.MINOR.PATCH).
• Keep version values synchronized across Python, Node, and docs files.
• Update CHANGELOG.md for user-visible changes.
• Do not tag releases from unvalidated branches.

PR checklist

• Keep changes scoped and minimal.
• Preserve cross-platform support (Windows/macOS/Linux).
• Keep Python and Node installers behaviorally consistent.
• Update docs when behavior changes.
• For security-related changes, explicitly document risk and mitigation in the PR description.

If you are unsure about contribution structure, use AGENTS.md as the authoritative development guide for coding agents and contributors.

Secure Publishing (Maintainers)

For npm CI/CD publishing, this repo now uses Trusted Publishing (OIDC) instead of long-lived write tokens.

1. On npm, open package settings for extension-agent-installer
2. Configure Trusted Publisher:
   • Provider: GitHub Actions
   • Organization/user: lmt-expert-company
   • Repository: extension-agent-installer
   • Workflow filename: publish.yml
3. Keep id-token: write permission enabled in .github/workflows/publish.yml

Notes:

• This avoids OTP/token security issues in CI
• NPM_TOKEN is no longer required for trusted publishing
• PyPI publishing still uses PYPI_API_TOKEN

License

MIT License - see LICENSE for details.

---

Disclaimer

IMPORTANT: Third-Party Security Tools Disclaimer

This skill relies on third-party security scanning tools and APIs developed and maintained by independent organizations:

1. mcp-scan is developed by Snyk - a third-party security company
2. Agent Trust Hub is operated by Gen Digital - a third-party security company

The author of this skill:

• Does NOT guarantee the accuracy, completeness, or reliability of security scans
• Is NOT responsible for any security issues that may not be detected by these third-party tools
• Is NOT affiliated with Snyk, Gen Digital, or any other third-party security providers
• Cannot be held liable for any damages resulting from the use of this skill or the third-party security tools it integrates

Users acknowledge that:

• Security scanning is provided by independent third parties
• Third-party tools may have limitations, bugs, or false positives/negatives
• Extensions may be modified after scanning
• Users should always review extension source code before installation
• Security tools may share data with their respective providers (see their privacy policies)

Third-Party Terms:

• Using mcp-scan means you agree to Snyk's Terms of Use and Privacy Policy
• Using Agent Trust Hub means you agree to Gen Digital's Terms and Privacy Policy

Use this skill at your own risk. Always verify extensions manually before installation.