68 skills · sorted by relevance
Design agents for multi-turn tool use — sequential tool calls, result accumulation, error recovery, and complex task decomposition over multiple turns.
The English summary is being prepared. Showing the original description.
OpenAI GPT Image generation and editing (gpt-image-2, 1.5, 1, mini). Text-to-image, mask-based inpainting, multi-reference composition, multi-turn conversational editing via Responses API, streaming with partial images. This skill should be used when generating or editing images via OpenAI's image models, when near-perfect text rendering in images is needed, when mask-based region-aware editing is required, when multi-turn conversational image editing is desired, or when streaming progressive image delivery is needed. Complements nano-banana-pro (Gemini) as a parallel image generation backend.
The English summary is being prepared. Showing the original description.
Design Agentforce conversations that span multiple turns without losing context: session variable scoping, conversation memory, clarifying-question patterns, topic-to-topic (now subagent) handoff, and the right abstractions for accumulating state across turns. NOT for deciding the topic boundaries themselves or out-of-scope behavior — use agentforce/agent-topic-design. NOT for single-turn agent actions and their input/output contracts — use agentforce/agent-actions.
The English summary is being prepared. Showing the original description.
Set up and run multi-turn RL training for interactive environments (terminal tasks, tool use, search/RAG, games) using the Tinker API. Use when the user wants multi-turn RL, agentic training, tool-use RL, or interactive environment training.
The English summary is being prepared. Showing the original description.
Set up reinforcement learning, training guided by response rewards, for large language models. Use Megatron-LM and SGLang to prepare data, configure training and response generation, and build asynchronous or multi-turn experiments.
Set up a realtime evaluation folder with a suitable shared harness, the framework that runs evaluations. Prepare prompts, tool definitions, realistic starter data, and a README, then validate the setup with evaluation and test runs.
Guide reinforcement learning, training models through rewards, with slime. Prepare data, configure Megatron-LM training and SGLang response generation, and set up asynchronous runs or custom multi-turn workflows for supported language models.
Set up LLM serving with structured responses, including JSON and outputs restricted to specified patterns. Covers reusing shared prompt calculations, tool calling, and multi-turn conversations through SGLang, with examples for server setup and application integration.
Build image-based conversations with LLaVA, an AI model that combines image understanding with language. Covers setup, model selection, image descriptions, follow-up questions, and memory-saving options for running models on a GPU.
Generate text and images through a reverse-engineered Gemini Web API, an interface reconstructed from the web service. Use reference images for descriptions or variations, and continue conversations across requests with saved sessions.
Build conversations and workflows with the Gemini Interactions API on Gemini Enterprise Agent Platform. Follow Python or JavaScript examples for server-stored conversation history, streaming responses, structured JSON output, function calls, and background agent tasks.
Build repeatable DeepEval tests for AI agents, chatbots, and applications that retrieve documents to answer questions. Reuse or generate evaluation datasets, inspect failures, and improve prompts, tools, or retrieval through repeated evaluation runs.
Build image conversations with LLaVA, a model that connects image recognition with language. This skill guides setup, image descriptions, visual questions, and follow-up conversations, with examples for Python, command-line use, and a Gradio interface.
Set up reinforcement learning, training guided by rewards, for language models with slime. The skill covers Megatron-LM training, SGLang response generation, data preparation, asynchronous execution, and custom workflows for agents that use tools across multiple turns.
Run concurrent Omnigent conversations with real hosts and runners while replacing the LLM with a zero-latency mock. Read the generated report to explain response-time distributions, throughput, and failures, helping assess Omnigent's own processing overhead.
Guide a workshop one question at a time with visible progress and clear answer options. The skill adapts to supplied context, handles interruptions and pauses, and closes with a summary of decisions and actionable choices.
LLM application red-teaming — prompt injection (direct + indirect), jailbreak, system-prompt leak, data exfiltration, guardrail bypass, multi-turn crescendo, cross-lingual + cipher + invisible-unicode token smuggling, excessive agency / tool abuse, insecure output handling. Canonical OWASP LLM Top 10 + ASI01-ASI10 mapping. Use when a target exposes a chat/completions/assistant/copilot endpoint, an AI feature that consumes user text or documents, or any /v1/chat, /api/chat, /mcp surface.
The English summary is being prepared. Showing the original description.
Use this skill when writing code that calls the Gemini API for text generation, multi-turn chat, multimodal understanding, image generation, video generation, speech generation (TTS), voice design, voice replication, streaming responses, background research tasks, function calling, structured output, or migrating from the old generateContent API. Covers SDK usage and best practices for Gemini models and agents in Python and TypeScript.
The English summary is being prepared. Showing the original description.
Use this skill when OpenStoryline is already installed and the user wants to start the local MCP/Web services, create or continue a session, send editing instructions, perform multi-turn re-editing, and verify rendered video outputs, as well as Chinese requests like “启动 OpenStoryline”, “把 OpenStoryline 跑起来”, “用 OpenStoryline 剪视频”.
The English summary is being prepared. Showing the original description.
Use this skill when a user wants to create, run, or analyze evaluation suites for Microsoft 365 Copilot declarative agents with the public @microsoft/m365-copilot-eval CLI. Trigger on intents such as "evaluate my agent", "test my agent", "run my evals", "create eval prompts", "add multi-turn tests", "tune evaluator thresholds", "why is my agent failing", or "set up eval environment variables".
The English summary is being prepared. Showing the original description.
Conducts multi-turn iterative deep research on specific topics within a codebase with zero tolerance for shallow analysis. Use when the user wants an in-depth investigation, needs to understand how something works across multiple files, or asks for comprehensive analysis of a specific system or pattern.
The English summary is being prepared. Showing the original description.
Train and run models on Tinker (Thinking Machines) — LoRA and full fine-tuning, SFT with correct loss masking, RL/RFT with group-relative advantages and importance sampling, agentic RL over multi-turn tool-using agents, inference via the SamplingClient or the OpenAI-compatible endpoint, checkpoint management and export, every hyperparameter, and cost estimation from the live price table. Use this skill whenever the user mentions Tinker, tinker://, TINKER_API_KEY, console.tinker.ai, or the tinker CLI; wants to fine-tune, SFT, RFT, or RL a model on Tinker; asks about Tinker models, context windows, LoRA rank, sampling params, loss functions (cross_entropy, importance_sampling, ppo, cispo, dro), Datum construction, forward_backward or optim_step; wants to sample from, download, resume, merge, or export a Tinker checkpoint; or asks what a Tinker run will cost. Whenever Tinker work is initialized in a project, this skill also creates a TINKER_PRICING.md there so prices sit next to the training code. Reach for it even on vague asks like "train a model on this data" or "get inference working" when Tinker is the platform in play.
The English summary is being prepared. Showing the original description.
Interactive interview that formalizes a fuzzy research idea into a structured spec (RQ, hypotheses, identification, data needs, empirical strategy). Use when user says "interview me", "help me think through this idea", "I have a half-baked idea", "formalize this into a project", "walk me through framing a study". Multi-turn Q&A; saves spec to disk. NOT for lit review (`/lit-review`) or ideation from scratch (`/research-ideation`).
The English summary is being prepared. Showing the original description.
Generate structured research questions, testable hypotheses, and candidate empirical strategies from a topic, phenomenon, or dataset description. Use when user says "give me research ideas on X", "brainstorm questions about Y", "what could I study with this data?", "I'm looking for a paper idea on...", "generate hypotheses for...". One-shot generation, not multi-turn. For idea-refinement use `/interview-me`.
The English summary is being prepared. Showing the original description.