Universal Agent Plugins & Skills Ecosystem

Welcome to the central Open Standard Agent repository. This repository functions as a universal capability registry, fundamentally designed for AI agents, by AI agents.

Project Overview

This is a strictly cross-platform (Windows, Mac, Ubuntu) library that serves as a universal deployment source for multiple IDEs and agent frameworks, including:

• Claude Code (.claude/)
• GitHub Copilot (.github/)
• Gemini CLI / Antigravity (.gemini/, .agent/)
• Roo Code, Windsurf, Cursor, and other compliant integrations.

Instead of keeping documentation trapped in disparate folders, this site serves as the navigable master hub for all active agent skills and capabilities.

---

Installation

Option 1: Claude Plugin Marketplace (Claude Code native)

This repository is a published Claude Code marketplace. Add it once and install any plugin by name:

/plugin marketplace add richfrem/agent-plugins-skills
/plugin install context-bundler
/plugin install spec-kitty
/plugin install rlm-factory

To browse and install interactively, run /plugin and go to the Discover tab — all 27 plugins from this marketplace will be listed there. Run /reload-plugins after installing to activate without restarting.

Option 2: npx skills CLI (Mac / Linux, cross-agent)

Use the npx skills CLI to install directly into any agent environment. It auto-detects installed agents (Claude Code, GitHub Copilot, Gemini, Cursor, and 30+ others) and wires everything up natively.

# Install all plugins from this repo
npx skills add richfrem/agent-plugins-skills

# Install a single plugin
npx skills add richfrem/agent-plugins-skills/plugins/rlm-factory
npx skills add richfrem/agent-plugins-skills/plugins/vector-db
npx skills add richfrem/agent-plugins-skills/plugins/spec-kitty-plugin

# Update all installed skills to latest
npx skills update

Warning

Windows users: do not use npx skills add with this repository. This repo uses Git symlinks inside skill directories. On Windows, Git checks out symlinks as plain text files containing the relative path string. The npx installer copies the text file as-is instead of following the symlink — resulting in broken one-line pointer files in .agents/ that Python cannot execute. Use the Bridge Installer (Option 3) instead.

Option 3: Bridge Installer (Windows — works everywhere)

The Bridge Installer is a pure-Python alternative to npx that understands Windows Git pointer files. It resolves symlink pointers at install time, writes real hard copies into .agents/, and creates Junctions (Windows) or symlinks (Mac/Linux) into each agent environment.

First install — run once from the cloned repo root:

python plugins/plugin-manager/scripts/install_all_plugins.py

Subsequent installs / consuming projects — once installed, invoke directly from .agents/:

# Reinstall / update all core plugins
python .agents/skills/bridge-plugin/scripts/install_all_plugins.py

# Add plugins from a second source without touching existing installs
python .agents/skills/bridge-plugin/scripts/install_all_plugins.py --plugins-dir path/to/other/plugins

# Preview without writing anything
python .agents/skills/bridge-plugin/scripts/install_all_plugins.py --dry-run

No Node.js required. No symlink permission issues. Installs are additive — running against multiple plugin directories merges them cleanly.

Installing Locally (For Contributors / Local Dev)

After cloning, install the full plugin ecosystem from source:

# Install everything
python plugins/plugin-manager/scripts/install_all_plugins.py

# Install a single plugin
python plugins/plugin-manager/scripts/bridge_installer.py --plugin plugins/rlm-factory

To update after making local changes, re-run the same command — the installer overwrites existing entries and skips any files currently locked by your IDE.

Browse all available plugins and their install status at: skills.sh/richfrem/agent-plugins-skills

---

Core Architecture: Agent Loops

The system relies heavily on the Agent Loops architecture, a sophisticated routing system that unifies state management across complex agent executions.

Agent Loops ensure shared closure across all operations, which means no session terminates without saving memory (persisting context snapshots), formalizing learnings (distillation), and sealing its actions. The architectural flows are categorized into 5 core behavior patterns:

1. Orchestrator: The intelligent task router and lifecycle administrator that initiates and concludes sessions.
2. Simple Learning Loop: The foundational pattern for research, contextual integration, and continuous memory persistence.
3. Red Team Review: An adversarial, multi-agent evaluation framework used to stress-test designs and catch architectural drift.
4. Dual-Loop: An inner execution / outer verification paradigm for complex, multi-step engineering assignments requiring ongoing correction.
5. Agent Swarm: A parallelized execution framework enabling concurrent sub-agents to process distinct tasks simultaneously on independent worktrees.

---

Site Navigation & Plugin Ecosystem

This repository currently hosts 29 distinct plugin packages containing over extensive modular agent skills. Below is the complete manifest of capabilities, each linked to its respective definition document (SKILL.md).

Architecture & Design

• ADR Manager — Automatically draft and syndicate Architecture Decision Records. ↳ adr-management

Agent Operations & Management

• Agent Scaffolders — Interactive creators to scaffold exact file hierarchies for plugins, skills, hooks, Azure API wrappers, GitHub workflows, and MCP configurations. ↳ audit-plugin ↳ create-agentic-workflow ↳ create-azure-agent ↳ create-github-action ↳ create-hook ↳ create-legacy-command ↳ create-mcp-integration ↳ create-plugin ↳ create-skill ↳ create-sub-agent
• Plugin Manager — The authoritative suite for ensuring ecosystem health, synchronization, garbage collection, and artifact bootstrapping. ↳ ecosystem-cleanup-sync ↳ plugin-bootstrap ↳ plugin-maintenance ↳ plugin-replicator
• Plugin Mapper — Intelligent transpiler that maps universal .md tasks to IDE-specific rule configurations (e.g. converting prompts for .claude/). ↳ agent-bridge

Standards & Hygiene

• Agent Skill Open Specifications — The canonical documentation for agents on how to construct compliant skills and directory architectures. ↳ ecosystem-authoritative-sources ↳ ecosystem-standards
• Coding Conventions — Centralized rules engine for standardizing file headers and snake_case/camelCase/PascalCase semantics across multiple languages. ↳ coding-conventions ↳ conventions-agent

Data & Code Manipulation

• Context Bundler — Tooling to package deep directory contexts and code traces into single readable files. ↳ context-bundling
• Dependency Management — Handles cross-platform pip-compile workflows securely. ↳ dependency-management

Agent Persona CLI Interfaces

Sub-systems allowing fresh, isolated model context spaces for advanced tasks (audits, QA, external validation).

• Claude CLI Agent ↳ claude-cli-agent
• Copilot CLI Agent ↳ copilot-cli-agent
• Gemini CLI Agent ↳ gemini-cli-agent

Artificial Constraints & Vectors

• HuggingFace Utils — Snapshot persistence and HuggingFace Soul repository lifecycle actions. ↳ hf-init ↳ hf-upload
• Memory Management — Orchestrates multi-tiered cognition and context caching between long-term persistent storage and active memory files. ↳ memory-management

Authoring & Documentation

• Doc Coauthoring — Iterative and collaborative tech specification drafting. ↳ doc-coauthoring
• Link Checker — Continuous markdown hyperlink validation. ↳ link-checker-agent
• Markdown to MSWord Converter — Interoperability translator for non-technical stakeholders. ↳ markdown-to-msword-converter
• Mermaid to PNG — Diagram exporter and renderer. ↳ convert-mermaid

Feature Driven Engineering

• Exploration Cycle — Autonomous discovery loop for idea framing, business requirements capture, user stories, prototyping, and handoff into formal engineering specs. Sits before and alongside the Spec-Kitty lifecycle. ↳ exploration-workflow ↳ exploration-session-brief ↳ business-requirements-capture ↳ user-story-capture ↳ exploration-handoff ↳ exploration-optimizer
• Spec Kitty Suite — The massive enterprise-grade Spec-Driven Development implementation layer handling the Spec -> Plan -> Tasks -> Implement -> Merge workflow autonomously. ↳ spec-kitty-accept ↳ spec-kitty-agent ↳ spec-kitty-analyze ↳ spec-kitty-checklist ↳ spec-kitty-clarify ↳ spec-kitty-constitution ↳ spec-kitty-dashboard ↳ spec-kitty-implement ↳ spec-kitty-init ↳ spec-kitty-merge ↳ spec-kitty-plan ↳ spec-kitty-research ↳ spec-kitty-review ↳ spec-kitty-specify ↳ spec-kitty-status ↳ spec-kitty-sync-plugin ↳ spec-kitty-tasks ↳ spec-kitty-update ↳ spec-kitty-workflow
• Task Manager — Kanban board synchronization and task lane transitions. ↳ task-agent

Knowledge Base / Vector Stores

• Obsidian Integration — Bi-directional sync translating standard codebase folders into fully operational Graph Vaults inside Obsidian. ↳ obsidian-bases-manager ↳ obsidian-canvas-architect ↳ obsidian-graph-traversal ↳ obsidian-init ↳ obsidian-markdown-mastery ↳ obsidian-vault-crud
• RLM Factory — Reverse Language Modeling engine generating high-speed offline functional cache representations of code. ↳ ollama-launch ↳ rlm-curator ↳ rlm-distill (agent) ↳ rlm-cleanup (agent) ↳ rlm-init ↳ tool-inventory-init
• Tool Inventory — Vector-search registry for all python scripts making them accessible to agents without hard-coded rules. ↳ tool-inventory
• Vector DB — ChromaDB driven continuous semantic embedding indexing module for native codebase integration search. ↳ vector-db-search (read-only skill) ↳ vdb-ingest (agent) ↳ vdb-cleanup (agent) ↳ vector-db-init ↳ vector-db-launch

Utilities

• Migration Utils — Standard scripts utilized during extensive codebase restructuring architectures. ↳ migration-utils

---

Generated autonomously via Agent Extensions.