135 skills · sorted by relevance
Training Details of the Second RL Stage 4 Experiment 5 Ethics and Safety Statement 6 Conclusion, Limitation, and Future Work 7 Author List A Backgroun
The English summary is being prepared. Showing the original description.
Orientation for custom (non-OpenAI) models running in the Codex app through the codex-router proxy. Explains that the app's native tools arrive as flattened codex_app__ and mcp__ names, that the router restores them so the app executes them, which companion skills to read before threads, browser, or computer-use work, and that a turn with no tool call ends the task. Use when the session uses a custom (non-OpenAI) model, for example deepseek-v4-flash or mimo-v2.5, when codex_app__ or mcp__ tool names appear in the tool list, when a tool result just arrived and more work remains, or when thread, browser, or computer-use work is requested.
The English summary is being prepared. Showing the original description.
Control local apps through Computer Use (the @oai/sky runtime) inside the Codex app. Use when the session uses a custom (non-OpenAI) model, for example deepseek-v4-flash or mimo-v2.5, and the user asks to control the computer, operate a desktop app's UI, use Safari or Chrome through computer use, click or type in an app, or take a screenshot of an app. Prefer purpose-built connectors, APIs, or CLIs when they exist.
The English summary is being prepared. Showing the original description.
Create, list, read, message, wait on, fork, rename, archive, and pin Codex threads (sidebar tasks), plus automations and app navigation, using the app-native codex_app tools. Use when the session uses a custom (non-OpenAI) model, for example deepseek-v4-flash or mimo-v2.5, and the user asks to create a thread or a new task or agent, list or read threads, send a message to a thread, wait for a thread, fork or rename a thread, archive or pin a thread, set up an automation or reminder, or open something in the Codex app.
The English summary is being prepared. Showing the original description.
Drive the Codex in-app browser (open, navigate, click, type, screenshot, read page state) through the app's own node_repl runtime. Use when the session uses a custom (non-OpenAI) model, for example deepseek-v4-flash or mimo-v2.5, and the user asks to use the in-app browser, open or navigate a page in it, test a local app in a browser, or click, type, or take a screenshot in the Codex browser panel.
The English summary is being prepared. Showing the original description.
Sample skill that proves a DeepSeek Harness bundle imported into Codewhale. Use only when the user is trying the dsh-sample plugin.
The English summary is being prepared. Showing the original description.
Use when user needs capabilities Claude lacks (image generation, real-time X/Twitter data) or explicitly requests external models ("blockrun", "use grok", "use gpt", "dall-e", "deepseek")
The English summary is being prepared. Showing the original description.
Provides PyTorch-native distributed LLM pretraining using torchtitan with 4D parallelism (FSDP2, TP, PP, CP). Use when pretraining Llama 3.1, DeepSeek V3, or custom models at scale from 8 to 512+ GPUs with Float8, torch.compile, and distributed checkpointing.
The English summary is being prepared. Showing the original description.
Use when adding or updating Spark AI provider models or their compact and preflight metrics, especially baidu-oneapi, Grok, Claude, GPT, DeepSeek, OneAPI, context windows, output budgets, prices, thinking maps, or transports.
The English summary is being prepared. Showing the original description.
A2AP — AI Agent互联互通的第一开放协议。6个LIVE能力（humanize_text/memory_evolve/skill_audit/news_briefing/agent_peering/custom_agent_dev）全部通过DeepSeek V4 Pro真机实现。公网已打通 frp-can.com:51831，支持HTTP Bridge & MCP桥接（87K⭐生态直接调用）。Hermes↔OpenClaw双向通知桥。TCP原始协议（非HTTP），默认端口9800。CC BY-SA 4.0。
The English summary is being prepared. Showing the original description.
Provides PyTorch-native distributed LLM pretraining using torchtitan with 4D parallelism (FSDP2, TP, PP, CP). Use when pretraining Llama 3.1, DeepSeek V3, or custom models at scale from 8 to 512+ GPUs with Float8, torch.compile, and distributed checkpointing.
The English summary is being prepared. Showing the original description.
Use when adding or updating Spark AI provider models or their compact and preflight metrics, especially baidu-oneapi, Grok, Claude, GPT, DeepSeek, OneAPI, context windows, output budgets, prices, thinking maps, or transports.
The English summary is being prepared. Showing the original description.
Use when user needs capabilities Claude lacks (image generation, real-time X/Twitter data) or explicitly requests external models ("blockrun", "use grok", "use gpt", "dall-e", "deepseek")
The English summary is being prepared. Showing the original description.
Smart LLM router — save 67% on inference costs. Routes every request to the cheapest capable model across 41 models from OpenAI, Anthropic, Google, DeepSeek, and xAI.
The English summary is being prepared. Showing the original description.
Unified LLM Gateway - One API for 70+ AI models. Route to GPT, Claude, Gemini, Qwen, Deepseek, Grok and more with a single API key.
The English summary is being prepared. Showing the original description.