角色动态 sprite / 桌宠皮肤 / atlas 状态行生产线：母图锚身份 → Frame A 首帧 → AI 视频生成（人工环节）→ 截帧确定性加工 → atlas row + 三道闸。 Use when: 做桌宠皮肤、给猫补新状态动画行、visible-cafe/F258 猫形象素材、 角色小循环动画（呼吸/走路/睡觉/挥手级）、把静态角色"做成动态的"。 Not for: 动画短片成片（用 anime-forge）、录屏/教程视频（用 video-forge）、 单张静态图/头像（用 image-generation）、UI 图标。 Output: 192×208 atlas row strip + QA 包（contact/GIF/provenance）→ 过三道闸 + operator identity veto 后接 pet.json。

Register Model Context Protocol servers on the host with sbx mcp and expose them to sandboxes through the built-in MCP gateway — registration, OAuth, static vs dynamic mode, and Cedar MCP policy. Use when adding an MCP server to a sandbox, deciding between --static-mcp and dynamic discovery, wiring OAuth credentials for a remote server, or writing a governance policy for MCP tool calls.

Locate the official Docker Sandboxes page behind a claim, build or refresh the local mirror of the docs, and carry what changed upstream into the other sbx skills. Use when checking whether an sbx claim still holds, when `references/docs/` is missing, when an sbx release lands, or when a skill's verification date needs moving.

Work out why a Docker Sandbox cannot reach a host, install a package, keep a file, or start at all. Use when a download or clone fails inside sbx, a credential is rejected, a sandbox is killed or stops on its own, or something written inside it disappeared.

Run and operate Docker Sandboxes — choosing an agent, passing it flags, git workspace modes, ports, secrets, network policy, and sharing host skills into a sandbox. Use when starting or re-attaching to a sandbox, picking between the built-in agents, signing in from CI, copying files in or out, or reaching for the right sbx command.

Multivariate severity assessment and humane endpoint prediction for laboratory animal studies using the RELSA (RELative Severity Assessment) score and ARIMA-based foRcast forecasting. Use when combining welfare readouts — body weight or weight loss, body temperature, clinical or nesting scores, biomarkers, activity, heart rate, burrowing, wheel running — into one severity score per animal per day, when asking which animals are at risk of reaching a humane endpoint or when one will be reached, when defining attention/danger zones or thresholds on a severity scale by kernel density estimation, or when reporting severity for a 3Rs, refinement, animal-welfare, or EU Directive 2010/63/EU severity-assessment context. Covers directionality ("turned" variables), baseline normalization, reference sets, RELSA weights, ARIMA prediction intervals, and RMSE/PICP/MPIW evaluation.

Read, search, summarize, and triage AgentMail inboxes through the connected MCP server. Use for ANY request to look at, search, or process mail — even a simple 'search my inbox for X' or 'any new mail?'; the read workflow applies regardless of task size. Also use to summarize conversations, inspect attachments, manage read/unread labels, or find messages needing a reply; do not use for sending or drafting (agentmail-send-email), inbox administration (agentmail-manage-inboxes), or MCP connection setup (agentmail-mcp).

Guide to the assistant-ui runtime system, the state and action layer behind every chat surface. Use when creating a runtime (useLocalRuntime with a ChatModelAdapter, useExternalStoreRuntime for Redux or Zustand backed stores, useRemoteThreadListRuntime, useCloudThreadListRuntime, useAssistantTransportRuntime), wiring AssistantRuntimeProvider, or reading and mutating thread, message, composer, and attachment state. Covers useAui, useAuiState, and useAuiEvent with the stable selector rule, the scope accessors (aui.thread, aui.message, aui.composer, and siblings), the AuiConfig plus AuiProvider grammar, RuntimeCapabilities, and the attachment, speech, dictation, feedback, suggestion, and history adapters. Route here for a provider error ('Cannot read property of undefined'), a removed v0.12 hook import, an infinite re-render from an object literal selector, or state not updating. For multi-thread list UI use thread-list instead; for toolkit definitions and tool call UI use tools instead.

Installs and configures assistant-ui with the `assistant-ui` CLI (`create`, `init`, `add`, `update`, `upgrade`, `codemod`, `doctor`, `info`, `mcp`, `agent`) and picks the runtime adapter for a backend. Use for first-time install (`npx assistant-ui@latest create`, templates, `--example` scaffolds), adding to an existing Next.js app (`init`), or keeping an install current. Covers the runtime hook per backend: `useChatRuntime` (AI SDK v7), `useLangGraphRuntime`, `useStreamRuntime` (LangChain), `useAdkRuntime`, `useA2ARuntime`, `useAgUiRuntime`, `useEveAgentRuntime`, `useOpenCodeRuntime`, `useLocalRuntime`, `useExternalStoreRuntime`, `useDataStreamRuntime`, and `useAssistantTransportRuntime`, plus wiring guides for Mastra, Cloudflare Agents, Claude Managed Agents, LLM gateways, and Electron, the Expo (`--native`) and React Ink (`--ink`) scaffolds, and the hosted MCP docs server. Once a runtime exists, use elements for the styled Thread, runtime for hooks and adapters, and update for a deep migration.

Draft, send, reply to, or forward email through the connected AgentMail MCP server. Use for ANY request to send, reply to, or forward mail — even a quick one-line send or a simple forward that looks like a single tool call; the sending rules apply regardless of task size. Also use to compose messages, create or schedule drafts for review, or send to named recipients; do not use for reading or triaging mail (agentmail-check-email), inbox administration (agentmail-manage-inboxes), or MCP connection setup (agentmail-mcp).

Installs and customizes assistant-ui elements, the styled shadcn-style component catalog served from the r.assistant-ui.com registry through `npx assistant-ui@latest add <item>`. Use when adding a prebuilt chat surface or widget (Thread, ThreadList, AssistantModal, AssistantSidebar, ToolFallback, ToolGroup, MarkdownText, Reasoning, Sources, Attachment, ModelSelector, Voice orb, McpConfig) or one of the 120 standalone elements (approval card, agent plan, code diff, data table, chart, and so on), choosing between runtime-connected `<name>.aui.tsx` files and props-driven standalone files, overriding Thread slots through the `components` prop, editing the copied source under `components/assistant-ui/elements/`, or picking the Radix versus Base UI flavor in `components.json`. Route here when an import from `@/components/assistant-ui/...` fails, an element renders unstyled, or the CLI installs the wrong flavor. For unstyled building blocks use primitives; for the CLI scaffold itself use setup.

Use sem for entity-level (function/class/method) code questions in any Git repo. Four questions, four verbs - where is it (sem find, sem grep), what does my change touch (sem impact), is it correct (sem check), what should a human review (sem certify). Trigger whenever the user asks where something is defined or called, what a change breaks or which tests to run, whether a change is correct, what changed in a commit or PR, or what to review; also use it proactively before refactors and in code review, where line-level git output is noisy.

Inspecting the user's current Positron configuration, read live from this running Positron -- state, not documentation: docs pages, release notes, and settings.json file reads cannot answer these for this install; this skill's commands can. Use for: what the user has configured, what a setting is set to and where, why a setting they set is not taking effect, which features are in preview or experimental and what enables each, whether a setting exists, its default and allowed values, and which AI model providers are enabled, signed in, or failing to authenticate. Triggers: "what settings do I have enabled", "which features are in preview", "why isn't this setting working", "what is X set to", "is there a setting for X", "which providers do I have enabled".

Run OpenMed models fully on-device with the MLX (Apple Silicon), CoreML (iOS/macOS), or ONNX/WebGPU (cross-platform/browser) backends, including convert-quantize-run workflows. Use when the user wants to deploy OpenMed at the edge, run NER/de-id on Apple Silicon, target iPhone/iPad/Mac, export to ONNX or WebGPU, quantize a clinical model to int8/4-bit, run with no network, or pick between MLX/CoreML/ONNX. Covers the mlx/coreml/onnx extras, the convert() functions and python -m convert CLIs, quantization, loading a local MLX artifact through analyze_text, OpenMedMLXLanguageModel/generate_text, and the on-device-only PHI guarantee (nothing leaves the host).

Score an OpenMed clinical or biomedical NER model against a user-supplied gold corpus with entity-level precision, recall, and F1, then break errors down per label. Use when the user wants a seqeval-style scorecard, strict vs partial (relaxed) span matching, a per-label confusion matrix, false-negative / false-positive examples, or to debug why a model misses entities. Trigger on "evaluate NER", "entity-level F1", "seqeval", "precision recall F1", "confusion matrix", "error analysis", "strict vs partial match", or "score against gold" in an OpenMed context. The gold corpus is user-supplied; OpenMed bundles no i2b2/n2c2/MIMIC data.

Canonical patterns for writing custom MSBuild targets. USE FOR: diagnosing and fixing custom target authoring anti-patterns; broken SDK target chains across files (e.g., Directory.Build.targets silently redefining SDK targets); targets that replace CompileDependsOn instead of extending it with $(CompileDependsOn); query targets returning stale results from Outputs vs Returns misuse; missing Inputs/Outputs causing unnecessary rebuilds; missing FileWrites registration. Covers DependsOnTargets vs BeforeTargets vs AfterTargets, the Build→CoreBuild three-level pattern, and the $(XxxDependsOn) chain-extension pattern. DO NOT USE FOR: incremental build tuning (use incremental-build), parallelization (use build-parallelism), general anti-patterns (use msbuild-antipatterns), non-MSBuild build systems.

Toggle peon-ping sound notifications on/off. Use when user wants to mute, unmute, pause, or resume peon sounds during a Claude Code session. Also handles config changes like volume, pack rotation, categories — any peon-ping setting.

Prepare and publish DeepChat releases in this repository. Use when Codex needs to bump the app version, update CHANGELOG.md, keep release notes bilingual from v1.0.1 onward with English bullets first and Chinese bullets second, run release checks, create or update versioned release branches such as release/v1.0.1, continue a half-finished release, fast-forward main with the documented release flow, create or push version tags, or clean up release branches after publishing.

Optimizes the performance of existing Liger Kernel Triton kernels. Profiles kernels, diagnoses bottlenecks (memory-bound vs compute-bound), generates multiple optimization variants with benchmarking, and applies the best variant while maintaining correctness. Supports GPU architecture-specific optimization (Ampere, Hopper, Blackwell). Use when a user asks to optimize, speed up, tune, profile, or reduce memory of an existing Liger kernel.

一个议题，一场圆桌：主持人请来 3-5 位真实人物，定义开场，逐轮交锋， 每轮收一张 ASCII 结构图，用户用指令控节奏（可/止/深入此节/引入新人物）， 散场后全文存入 org 笔记。Use when user says "圆桌讨论", "圆桌", "roundtable", "辩论", or wants to explore a topic through multi-perspective structured debate.