jdiegosierra-enterprise-agent-plugins

Enterprise Agent Plugins

A scaffold for building multi-platform AI coding agent plugins that work across Claude Code, opencode, and OpenClaw. Use this as a starting point for your own enterprise engineering plugins.

This repository demonstrates a production-ready plugin architecture where src/ is the single source of truth, and each AI platform reads from it via symlinks.

One repo serves as a single catalog for the entire organization. Each department owns its own plugin — employees install only the ones they need.

Plugins

Plugin	Department	Status
acme-engineering	Engineering	Full example (agents, skills, runbooks, hooks)
acme-people	People/HR	Placeholder — ready to customize
acme-finance	Finance	Placeholder — ready to customize
acme-marketing	Marketing	Placeholder — ready to customize

Each plugin is independent — its own agents, skills, commands, and versioning. Employees from different departments install different plugins from the same marketplace.

Quick start

Pick your platform — you only need to install on the one(s) you use.

Claude Code

# From a Claude Code session:
/plugin marketplace add jdiegosierra/enterprise-agent-plugins
/plugin install acme-engineering@jdiegosierra-enterprise-agent-plugins

Updates are automatic — Claude Code checks for new versions on each session start.

Test locally without installing:

claude --plugin-dir ./plugins/acme-engineering/claude

opencode

# 1. Clone the repo
git clone https://github.com/jdiegosierra/enterprise-agent-plugins.git ~/repos/enterprise-agent-plugins

# 2. Run the setup script to symlink agents into opencode's config
bash ~/repos/enterprise-agent-plugins/plugins/acme-engineering/scripts/setup-opencode-agents.sh

# 3. Verify agents are registered
ls -la ~/.config/opencode/agents/

Updates: git pull in the repo and re-run the setup script.

OpenClaw

OpenClaw loads skills via extraDirs in openclaw.json. Point it to the src/skills/ directory (not openclaw/skills/ — OpenClaw doesn't resolve symlinks inside extraDirs).

# 1. Clone the repo into the OpenClaw workspace
git clone https://github.com/jdiegosierra/enterprise-agent-plugins.git /path/to/repos/enterprise-agent-plugins

# 2. Add to openclaw.json
cat <<EOF >> /dev/null
Add this to your openclaw.json under "skills.load":
{
  "skills": {
    "load": {
      "extraDirs": [
        "/path/to/repos/enterprise-agent-plugins/plugins/acme-engineering/src/skills"
      ],
      "watch": true
    }
  }
}
EOF

# 3. Restart OpenClaw (or wait — watch mode picks up changes automatically)

Updates: git pull in the repo. With watch: true, OpenClaw detects changes automatically.

What's included

The acme-engineering plugin is a fully functional example with:

• 2 agents — SRE (infrastructure, runbooks) and backend-developer (PRs, testing)
• 4 commands — help, setup, lint-fix, run-tests
• 3 skills — platform map, Kubernetes best practices, SRE runbook executor
• 1 runbook — GitHub access provisioning (end-to-end: Jira ticket to PR)
• 6 hooks — bash safety guard, welcome message, session rules, update checker, onboarding, notifications
• Multi-platform support — Claude Code, opencode, and OpenClaw with symlinks

Architecture