Skills — le catalogue de compétences de vos agents

Analyzes an uploaded photograph, provides detailed feedback on exposure, focus, composition, lighting, and suggests concrete camera settings and shooting technique. Generates a simulated improved image based on the recommendations.

对上传的艺术作品进行评估，提供点评、改进建议、模拟优化图，并列出该类作品的代表艺术家及代表作品

Rebuild photos as psychological or concept art

Use this skill when the user needs help with pure water and ultra-pure water system design, installation, commissioning, and operations. Triggers include: any mention of pure water system design, ultra-pure water, RO sys, EDI sys, water treatment process, membrane filtration, ion exchange, other

Build polished Wallpaper Engine music web wallpapers

文生图与图片编辑，外加电商场景全套：抠图、商品主图/场景图/细节图、活动海报、去水印消除、高清放大。

Confirms Sentinel's PII redaction guardrail survives routing through a modality-aware LiteLLM router to two different self-hosted model types -- a text model (Ollama) and a vision model (SmolVLM-Instruct) -- picked automatically by whether the request contains an image. Checks both redaction and rou

Design and implement intentional web motion with GSAP, React, Next.js, ScrollTrigger, Flip, SplitText, accessibility, performance, and responsive behavior. Use when animation or interaction polish is part of a frontend task.

Build an original short-drama score pack of 8 to 15 instrumental beds for tension, romance, comedy, and tearjerker scenes. This short drama BGM studio and episode score generator writes loop-friendly short-drama music you own and can use commercially, then delivers a labeled short drama score you can audition before the cut. Use it as a short drama soundtrack for short-drama BGM, vertical drama music, episode score, romance beds, tension beds, and original short-drama music.

Build an original livestream BGM playlist of 10 to 20 instrumental tracks for a room, store, or live-commerce shift. This livestream background music pack and store-loop music generator writes waiting, chat, product, hype, and night beds as original livestream music you own and can use commercially, then delivers a labeled playlist you can audition before going live. Use it as a playlist BGM generator for livestream BGM, live commerce BGM, store loop music, waiting room music, in-store background music, and original stream music.

changelog: Enhanced with ClawHub metadata, expanded examples, and rationalization prevention

视频分析处理 — 本地视频反编译分析工具。将视频拆解为时间轴剧本、语音转文字、场景分析、跨模态关联和精华摘要，支持多ASR引擎切换（Whisper/Paraformer/SenseVoice）、中文NLP增强、PaddleOCR中文识别。v4.0 新增短视频平台适配（抖音/快手/B站/视频号）和自动剪辑建议（高光检测/冗余标记/EDL导出/字幕样式）。v4.1 新增tiny模型优先体验（75MB低门槛）、说话人分离质量评分、剪映draft.json导出。v4.2 新增场景管理（detect→slice一条链）、短视频爆款预测、实时直播分析（流式ASR+敏感词检测）。v4.3 新增纯音频输入（mp3/m4a/wav播客与录音）、批量队列（SQLite+硬件档位并发）、GPU自动加速（CT2 int8量化）、ASR配置统一（--asr-engine单参数）。

Reach for this for quick YouTube lookups: a pasted link or video ID, a fast summary, what a channel posted lately, a topic search. Transcripts, search and channel-latest in one small skill. Skip it for uploads and account chores.

Reach for this when a video needs to become text: a pasted YouTube link or ID, a transcribe/summarize/translate request, mining a video for facts, or a bare URL shared with 'what does this say?'. Skip it for uploads and account chores.

Reach for this when a YouTube playlist is in play: a pasted playlist link or PL... id, listing a course or series, finding one video inside a long playlist, or turning a whole playlist into transcripts. Skip it for creating playlists or account chores.

王者荣耀出装与节奏建议。当用户说在玩某个英雄、对面有难搞/克制自己的英雄、报当前装备并问怎么出装、出什么克制对面，或问野怪/BUFF/暴君主宰刷新时间与节奏时使用。基于本地装备、英雄与野区数据给出下一件装备、成装路线和资源节奏。

Schedule MRP production batches in Odoo (BaconCo) — the fluid shop-floor scheduler. Periodically read open `mrp.production.batch` records, rank them into a run order by the shop rule (manual pins first, then art readiness, then imminent event dates, then earliest governing deadline), write that order back to the floor, and refine it into machine + time slots where the picture is solid. Defers batches whose decals aren't printed or whose artwork isn't digitized below work that can actually run, and expedites them to the top once they're about to be late. Reads purchase orders + vendor tracking

Use for low-budget Chinese 3D animation artwork (土味3D/国产低成本动画/early 2000s computer animation look). Triggers on: 低成本3D动画, 土味3D, 国产动画风格, low-budget 3D, rough computer graphics, amateur 3D animation, cheap 3D cartoon. Excludes polished Western animation studios, modern low-poly art, anime.

Track houses for sale on a private map. Use when the user texts a photo or screenshot of a property listing, pastes a listing URL, asks to add or remove a property, wants to rate one, leave notes on one, mark it toured or passed, refresh its price, or asks which houses they have saved. Looks up the full listing in the browser, keeps everything in ~/Plow/properties, and renders a map with photo pins.

How the Strategist works with the user to turn their own trading idea into explicit rules, backtest it honestly on Hyperliquid candle and funding history, and paper-trade it on testnet through the desk lifecycle. Method only - the desk ships no strategies and makes no return claims. Use when the user wants to design, test, compare or paper-trade an idea.

KYC media labeling Skill and KYC image labeling Skill for selected portrait, behavior, and scene review labels

Generate textual descriptions of one or more images when the current session model (e.g. opencode/big-pickle) has no image input, by delegating to a vision-capable model on OpenRouter via `opencode run` with --file. Use when the user drops photos into imgs/, asks "can you read/describe these images", or an image needs a description for an article, portfolio, infographic, or video pipeline. Related terms: 看图, 描述图片, image description, vision model, glm-4.6v, --file.

Analyze instrumental practice recordings to detect timing accuracy, pitch stability, tempo consistency, and dynamic range. Generates practice reports with targeted exercise recommendations. Use when practicing an instrument and wanting objective feedback on performance.

因果世界模型主动实验（Active Causal Experiment Design）。面对多个候选因果结构时，不被动观察，而是主动设计干预 do-operation，用期望信息增益/最优实验设计挑选最能区分候选结构的实验，Bayesian 更新后验，以最少实验辨识真实因果结构。当需要主动做实验以最快弄清变量间因果关系、设计 A/B 或干预实验、在多个竞争因果假设中高效辨识时使用。