lychee

Agent Extensions

Curated reusable agent skills and agents packaged as Claude Code plugins. Each subject (a tool, library, language, or workflow) is one plugin: skills are invoked as /<subject>:<facet> (e.g. /go:secure, /gh:send-pr) and agents are delegated as subagents. Canonical skills live under skills/ and agents under agents/; each subject is published as a self-contained plugin under plugins/, listed in the repo-root marketplace manifest (.claude-plugin/marketplace.json).

Claude Code is the only publication target.

Installation

Claude Code

# Add the marketplace (once)
/plugin marketplace add nq-rdl/agent-extensions

# Install a single subject
/plugin install go@rdl-agent-extensions

# Onboarding: install the rdl-team plugin
/plugin install rdl-team@rdl-agent-extensions

See docs/bundles.md for the full subject list.

Agent File Management

This repo keeps a single source of truth for top-level agent context files:

• AGENTS.md — single source of truth for agent contributor guidance.
• CLAUDE.md → symlink to AGENTS.md. Claude Code loads CLAUDE.md as project context; symlinking keeps the two in sync.

License

This repo is scope-licensed (not an OR dual-license — the license depends on the file, not the user's choice):

• Software — SPDX-License-Identifier: MIT. Full text: LICENSE.
• Media — SPDX-License-Identifier: CC-BY-4.0. Full text: LICENSE-CC-BY-4.0.