Advanced pattern for Ralph Loop using sub-agents, multiple models, and adversarial verification.

> ## Documentation Index

- [Agent CRUD](#agent-crud)

Use this package when the user wants persuasive marketing copy such as ads, landing page copy, launch messaging, headlines, or CTA variants.

Parallel task execution via Agent tool dispatches with worktree isolation. No experimental env var needed.

- [Agents Overview](https://raw.githubusercontent.com/google/adk-docs/refs/heads/main/docs/agents/index.md)

Async agents in tmux sessions with optional git worktree isolation.

{{PROJECT_ONE_LINE}}

`AGENTS.md` is the repo-level contract for agents working in this repository.

**Version 1.0.0**

This is your workspace. Treat it that way.

These are commonly used system-provided agents. This is not an exhaustive list — to see all

> Universal AI agent instructions for the DotRun (`dr`) CLI tool.

<!--

<!-- Managed by agent: keep sections & order; edit content, not structure. Last updated: 2026-01-14 -->

**Purpose**: This file provides coding conventions and operational commands for AI agents working in this repository.

If `BOOTSTRAP.md` exists in this workspace, follow it first. After completing its instructions successfully, delete it. Once deleted, you will not see it again — your ongoing instructions come from HE

This document defines the agent orchestration for the Procurement role. Agents work together to handle the full procurement lifecycle: vendor evaluation, price comparison, bid analysis, purchase order

This folder is home. Treat it that way.

Use this file only in stages `3.2` and `3.3`.

Agent definitions used by the /mz:plan skill. The skill reads these agent files, resolves the model from config, updates the agent frontmatter, and spawns via registered subagent names.

**角色**：Master Executor + 流程協調者

Custom agents provide the ultimate flexibility in ADK, allowing you to define **arbitrary orchestration logic** by inheriting directly from `BaseAgent` and implementing your own control flow. This goe

A2A is an open protocol from Google that defines how AI agents communicate, delegate tasks, and collaborate with each other. While MCP focuses on tool integration, A2A addresses agent-to-agent interop

'Run web, multi-source, or last-30-days research through AIsa. Use when: the user needs search, synthesis, competitor scans, or trend discovery. Supports research-ready outputs and structured retrieval.'

'Run web, multi-source, or last-30-days research through AIsa. Use when: the user needs search, synthesis, competitor scans, or trend discovery. Supports research-ready outputs and structured retrieval.'

'Run web, multi-source, or last-30-days research through AIsa. Use when: the user needs search, synthesis, competitor scans, or trend discovery. Supports research-ready outputs and structured retrieval.'

'Run web, multi-source, or last-30-days research through AIsa. Use when: the user needs search, synthesis, competitor scans, or trend discovery. Supports research-ready outputs and structured retrieval.'

Use loom workspace to create/manage worktrees for isolated execution of tickets.

maker-orchestrator

Manages parallel conversation branches using git worktrees. Use when you want to explore multiple solution approaches simultaneously, need to try different implementations, or want to checkpoint your current conversation state before branching into alternative paths.

|-

Scans and reports complete marketplace inventory (bundles, agents, commands, skills, scripts)

Automates updates to the Functional Area Master Index when new features are added or documentation changes. Maintains the single navigation source for all feature documentation, ensuring discoverability and organization.

Create, modify, and manage MCP tools for the Naki mahjong AI assistant. Use when adding new MCP tools, modifying existing tools, or fixing tool-related issues. This skill understands the Protocol-based MCPTool architecture.

Match SAE features between Ultra (24K) and Full (2K) models based on activation patterns and token overlap

Resolve common merge conflicts in merge-queue worktree including observations.jsonl, uv.lock, pyproject.toml, and egg-info files

methodology

methodology-system

mgrep is a semantic search tool that enables natural language queries across code, text, PDFs, and images. It is particularly effective for exploring larger or complex codebases where traditional patt

Launch ralph.sh in background. Best for large features (10+ tasks). Automatic execution until completion.

You are an expert conversion copywriter. Your goal is to write marketing copy that is clear, compelling, and drives action.

Evaluate solutions through multi-round debate between independent judges until consensus

Research a topic from the last 30 days on Reddit + X + Web, become an expert, and write copy-paste-ready prompts for the user's target tool.

Search, create, and manage notes in the Obsidian vault with wikilinks and index notes. Use when user wants to find, create, or organize notes in Obsidian.

Campaign attribution analysis involves systematically evaluating evidence to determine which threat actor or

'Analyzes UEFI bootkit persistence mechanisms including firmware implants in SPI flash, EFI System Partition

Configure Google Workspace advanced phishing and malware protection settings including pre-delivery scanning,