Loading...
Loading...
Found 16 Skills
Adopt multiple expert personas sequentially for complex problem analysis from diverse perspectives. Single-agent only — do NOT spawn sub-agents.
Production-ready reinforcement learning algorithms (PPO, SAC, DQN, TD3, DDPG, A2C) with scikit-learn-like API. Use for standard RL experiments, quick prototyping, and well-documented algorithm implementations. Best for single-agent RL with Gymnasium environments. For high-performance parallel training, multi-agent systems, or custom vectorized environments, use pufferlib instead.
Master skill for parallel subagent-driven execution with automatic fallback to single-agent sequential mode. Use when implementing plans with multiple independent sub-phases (SP1, SP2...) to dispatch parallel subagents, or when requiring code review between implementation and testing.
Build single-agent and multi-agent systems using Google's Agent Development Kit (ADK) in Python, Java, Go, or TypeScript. Use when creating AI agents with ADK, designing multi-agent architectures, implementing agent tools, configuring agent callbacks, managing agent state, orchestrating sequential/parallel/loop agent workflows, or when the user mentions ADK, google-adk, google agent development kit, agentic AI with Gemini, or agent orchestration with Google tools. Also use when setting up ADK projects, writing agent tests, deploying agents, or integrating MCP tools with ADK.
Byzantine fault-tolerant consensus and distributed coordination. Queen-led hierarchical swarm management with multiple consensus strategies. Use when: distributed coordination, fault-tolerant operations, multi-agent consensus, collective decision making. Skip when: single-agent tasks, simple operations, local-only work.
Claims-based authorization for agents and operations. Grant, revoke, and verify permissions for secure multi-agent coordination. Use when: permission management, access control, secure operations, authorization checks. Skip when: open access, no security requirements, single-agent local work.
Design multi-agent architectures for complex tasks. Use when single-agent context limits are exceeded, when tasks decompose naturally into subtasks, or when specializing agents improves quality.
Coordinate Claude Code Agent Teams through filesystem-based protocol. Use when orchestrating multiple Claude agents on parallel tasks, need task dependency management, multi-agent code review or implementation. Do not use when single-agent work suffices, task is not parallelizable.
Agent spawning, lifecycle management, and coordination patterns. Manages 60+ agent types with specialized capabilities. Use when: spawning agents, coordinating multi-agent tasks, managing agent pools. Skip when: single-agent work, no coordination needed.
Guides engineering of multi-agent systems—agent roles and specialization, orchestration topologies (supervisor, peer-to-peer, hierarchical, blackboard), task decomposition and routing, inter-agent messaging (A2A-style patterns), shared vs partitioned state, fan-out/fan-in and DAG workflows, synchronization and consensus, conflict resolution, fault tolerance and retries across agents, cost/latency/token budgets, cross-agent observability, testing multi-agent flows, and deployment (queues, durable workflows). Framework-agnostic; high-level LangGraph, Deep Agents, and agenthub—not single-agent loops (agentic-ai-developer), ML training (ai-engineer), strategy-only whiteboard (enterprise-strategist), or PM planning (technical-program-manager). Use for multi-agent system, multi-agent engineer, agent orchestration, supervisor agent, agent topology, fan-out fan-in, agent handoff protocol, multi-agent workflow, agent coordination, blackboard pattern, hierarchical agents, A2A, agent DAG, multi-agent architecture.
Spawns an Agent Team to collaboratively plan Power Platform / Dataverse applications. Three specialists (Data Architect, UX Designer, The Skeptic) debate and refine the plan before any code is written. Falls back to structured single-agent planning if agent teams are not enabled. Triggers on: "plan my app", "plan with team", "design my app", "architect this app", "plan the schema", "team planning", "agent team plan", "plan power app", "plan dataverse app", "design the data model".
This skill should be used when the user wants to review code, audit a diff, get a second opinion on changes, or run an adversarial review of files in the current working tree. Common triggers include "review this code", "audit this diff", "find issues in", "second opinion on this", "harsh review of", "adversarial review", and "security review of". Picks one or more reviewer personas (adversarial, security, architecture, performance). Reviews local files, `git diff`, or `git diff --staged` only — does not fetch external content. Runs in one of four modes: single-agent (one persona in the current agent), cross-model handoff (independent second opinion via another local AI CLI, with secret-shield preflight + prompt-shield wrap), multi-bg-agent (one persona per parallel background subagent), or agent-team (Claude Code Teams or equivalent on supporting agents). Skip when the user wants formatting fixes (use a linter) or refactoring patterns (use ts-best-practices or ts-best-practices-functional).