Total 55,942 skills, AI & Machine Learning has 9312 skills
Showing 12 of 9312 skills
Interactive QA session where users report bugs or issues through conversation, and the agent creates GitHub issues. Explore the codebase in the background to obtain context and domain language. Use when user wants to report bugs, do QA, file issues conversationally, or mentions "QA session".
Turn a vague, messy, or multi-part user ask into a clean, self-contained prompt that a fresh agent could execute without further questions. Interview the user one question at a time — walking down the decision tree, branching on each answer — until the prompt is tight, then output the final prompt as the deliverable. Trigger eagerly: any voice-dictated input, filler-heavy prose, underspecified references ("the thing", "that script"), multi-part requests, or any plan the user wants stress-tested. The skill itself can be skipped for trivial one-line requests where producing a prompt artifact would be pure ceremony — but once invoked, always produce the prompt, even if execution looks trivial.
Validate and use packed sequences and long-context training in Megatron-Bridge, distinguishing offline packed SFT for LLMs from in-batch packing for VLMs, and applying the right CP constraints.
Convert single-node scripts to multi-node Slurm sbatch jobs and debug common multi-node failures. Covers srun-native vs uv run torch.distributed approaches, container setup, NCCL timeouts, OOM sizing for MoE models, and interactive allocation.
Validate and use CPU offloading in Megatron Bridge, including layer-level activation offloading and fractional optimizer state offloading with HybridDeviceOptimizer.
Systematic workflow for MoE training optimization in Megatron Bridge, based on the Megatron-Core MoE paper. Covers the Three Walls framework, parallel folding, recompute strategy, dispatcher choice, and CUDA-graph bring-up.
This spell is archaeology, not history research. It operates on specific artifacts from a specific dead system to answer specific questions. It is NOT general tech history, NOT interviewing living people, and NOT monitoring live systems.
Compress an agent's routing file (RESOLVER.md or AGENTS.md) by converting granular skill-per-row tables into functional-area dispatchers. Each area lists sub-skills in a "(dispatcher for: ...)" clause. The LLM reads one area entry and routes to the correct sub-skill. Proven via held-out A/B eval: dispatcher pattern outperforms naive pipe-table compression.
Always-on ambient signal capture. Fires on every inbound message to detect original thinking and entity mentions. Spawn as a cheap sub-agent in parallel, never block the main response.
Evolve your brain's schema pack. Add page types, propose new ones from corpus scans, backfill page.type on existing pages, audit pack health. Triggers when an agent notices untyped pages, custom domains needing typed entities (researcher, contract, deposition), or wants to see what types the pack declares.
Use when starting any conversation — establishes how to find and use OAC skills, requiring Skill tool invocation BEFORE ANY response including clarifying questions, this is your secret weapon to best perform your tasks
Generate Chinese / Japanese speech with StepFun's stepaudio-2.5-tts — Contextual TTS that replaces step-tts-2's `voice_label` with natural-language `instruction` (≤200 chars) plus inline `()` parentheses for句内 prosody. Use when the user wants emotional / prosody control over voice synthesis (whisper, pause, stress, mood pivot mid-sentence), batch-generates game / app voice lines, migrates from `step-tts-2` (the `voice_label → instruction` breaking change), or hits StepFun's stricter 2.5-era censorship (死/消失/political terms). Triggers on 阶跃 TTS, StepAudio 合成, 语音合成, 配音, 文本转语音, TTS 升级, 迁移 step-tts-2. For transcription with the sibling stepaudio-2.5-asr model, use the stepfun-asr skill instead.