Browse reusable Agent Skills, each with a clear purpose and practical guidance.
High-performance reinforcement learning framework optimized for speed and scale. Use when you need fast parallel training, vectorized environments, multi-agent systems, or integration with game environments (Atari, Procgen, NetHack). Achieves 2-10x speedups over standard implementations. For quick prototyping or standard algorithm implementations with extensive documentation, use stable-baselines3 instead.
Run a full-sweep consolidation over all .reflexio/ files — TTL sweep + n-way cluster merge. Use when the user asks to 'clean up reflexio', 'consolidate memory', 'deduplicate playbooks', or suspects drift across sessions.
Captures user facts and procedural corrections into .reflexio/ so the agent learns across sessions. Use when: (1) user states a preference, fact, config, or constraint; (2) user corrects the agent and confirms the fix with an explicit 'good'/'perfect' or by moving on without re-correcting for 1-2 turns; (3) at start of a user turn, to retrieve relevant facts and playbooks from past sessions.
生成安全审计 skill。两种模式：(1) 项目模式——根据项目文档生成定制化审计 skill；(2) 通用模式——仅指定语言+框架，从参考资料库生成通用审计 skill。当用户想创建安全审计 skill、生成审计规则、或提到"生成安全审计skill"、"创建code review skill"、"生成 Java 审计 skill"时使用。
Text-based game suite for LLM training and evaluation (TextArena). 100+ single-player, two-player, and multi-player text games with an OpenAI Gym-style interface. Designed for benchmarking, self-play, multi-agent RL, and reasoning-focused LLM evaluation. Use for text-game environments, self-play RL, strategic reasoning evaluation, and agent-vs-agent tournaments.
To author, register, and test Rosetta hooks, add a SemanticKind, or debug a hook that won't fire.
To author, adapt, review, and validate prompts (skills, agents, workflows, rules, etc.) with brief, contracts, and a validation pack.
CRITICAL. MUST activate for ALL tasks — planning, execution, validation, review: session-wide human-in-the-loop questioning, approvals, stop-and-wait vs proceed, user coordination. NEVER assume approval. MANDATORY unless user requested EXACTLY `fully autonomous` or `No HITL`.