Media Skills

Use for website performance planning, Lighthouse/web-vitals checks, bundle/media/font budgets, Core Web Vitals triage, image/video optimization, animation performance, deployment readiness, and premium-site speed audits.

Direct website imagery, video, generative visuals, product renders, screenshots, gallery crops, hero media, thumbnails, and visual assets. Use when a site needs real or generated media rather than generic gradients, stock-like backgrounds, placeholder images, or uninspected crops.

Compose UI screenshots inside website layouts. Use for product sections, feature blocks, app showcases, case studies, dashboards, mockup stacks, detail crops, callout overlays, annotation systems, responsive screenshot art direction, and avoiding generic SaaS screenshot-card sections.

Use for premium frontend, website, app UI, landing page, dashboard, redesign, image-to-code, or visual-polish work that needs clearer design intent, adjustable visual direction, motion discipline, less generic output, and a final preflight check before handoff.

Prompt generated imagery for premium websites. Use when Codex needs hero images, product scenes, editorial visuals, backgrounds, portraits, venue imagery, campaign art, texture assets, mockups, or image-generation direction for luxury, cinematic, Squarespace-polished, Viktor Oddy-adjacent, or non-generic site visuals.

>- Plan and generate controllable Seedance video using Seedream 5.0 Pro storyboards and Seedance 2.0 today, with a Seedance 2.5 route when available. Use for consistent people, products, objects, food, or scenes; storyboard-to- video; reference-to-video; first-and-last-frame image-to-video; extensions; and Atlas Cloud media generation.

>- Write one model-agnostic video prompt spec, then compile it to whichever video model you can actually call. Use for cross-model prompt work, model comparison matrices, reusing one brief across providers, or when the target model is not yet available and the work must proceed on another one.

Agent-callable ElevenLabs tools — generate spoken audio from text, create sound effects and multi-speaker dialogue, re-voice and clean up audio, transcribe audio and video, design synthetic voices, and manage voices, history, and quota. Use when the user mentions ElevenLabs or wants AI audio work — text to speech, narration, voiceover, transcription, voice changing, or sound design — even if they don't name ElevenLabs explicitly, e.g. "read this aloud", "make an MP3 of this", "transcribe this recording".

Agent-callable HeyGen tools — generate AI avatar videos, translate and lip-sync videos, synthesize speech, and clone or browse voices and avatars. Use when the user mentions HeyGen or wants to create, translate, or check the status of AI videos, or to generate or clone a voice — including requests that don't name HeyGen, e.g. make an avatar video from a script, translate a video to Spanish, turn a script into a voiceover, or clone a voice from a recording.

Agent-callable Google Docs tools — create, read, export, and find documents; edit text, formatting, images, and page style. Use when the user mentions Google Docs or wants to create, read, search, or edit document content, even if they don't name Google Docs explicitly.

Agent-callable YouTube tools — search and read videos, update and delete videos, manage playlists and playlist items, read and post comments, rate videos, manage subscriptions, and read channel and caption metadata. Use when the user mentions YouTube or wants to find, comment on, or organize YouTube videos and playlists, even if they don't name YouTube explicitly.

Agent-callable Runway tools — generate images and video from text or images, edit and upscale media, animate characters, generate speech and sound effects, and run marketing recipes, plus track generation jobs. Use when the user mentions Runway or wants to generate, edit, or upscale AI media, even if they don't name Runway explicitly.

Use when the user has a video + an SRT and wants the subtitles either burned into the pixels (libass, always-visible) or soft-muxed as a togglable track. Also handles the final composite step for the localization pipeline — burn subs, mix a dub track, and keep the original audio as a low-volume bed, all in ONE ffmpeg encode (no cascade). Verifies libass availability and auto-downloads a static evermeet ffmpeg build when Homebrew's stripped binary lacks it. Triggers — "烧字幕", "硬字幕", "burn subtitles", "burn-in subs", "embed subtitle", "soft mux SRT", "把字幕烧进视频", "做最终合成".

Use when the user wants a 王建硕-style WeChat article (article.md) turned into a narrated short MP4 video — TTS voiceover via 火山引擎 Volcano TTS, HyperFrames CSS/GSAP animation per scene, subtle SFX, abstract watercolor background, full pipeline rendering to 1080×1920 portrait MP4 (30-90s). Triggers — "把这篇文章做成视频", "做一个解说视频", "讲解视频", "/wjs-converting-text-to-video".

Use when the user has a video + a target-language SRT and wants the video to actually speak that language — generates a time-aligned TTS voice dub. Routes by voice ID — Volcano (豆包) TTS for Chinese, edge-tts neural for any language. Defaults to one voice (single-speaker); opt-in multi-speaker via visual diarization. Outputs `*_<lang>_dub.mp4` with the dub audio in place of the original. Final mixing (audio bed + burn-in) is handed off to `/wjs-burning-subtitles`. Triggers — "配音", "中文配音", "Chinese dub", "voice over this", "dub the video", "TTS this SRT", "different voice for each speaker".

Use when migrating a WordPress site to a Hugo static site on GitHub Pages from a WXR export (.xml) plus the wp-content/uploads folder — preserving /archives/<id>/ URLs, localizing images, and deploying via GitHub Actions. Triggers — "把 WordPress 迁成 Hugo", "wordpress 转静态站", "migrate WordPress to Hugo", "WXR to Hugo", "publish WordPress to GitHub Pages", "/wjs-converting-wp-to-hugo".

Use when the user wants to capture a writer's voice from a few sample articles and rewrite text to sound like that author — "蒸馏文风", "提炼XX的风格", "把这篇改成XX的味儿", "学某某笔法改写", "mimic this writer", "match this author's style", "style transfer". Works for any author, 王建硕 included.

Use when the user has 2+ recordings of the same event (each with a `.sync.json` sidecar from wjs-syncing-multicam) and wants them combined into a single MP4 — auto-switching between cams second-by-second on audio energy, with optional picture-in-picture inset. Triggers — "auto-edit multicam", "做个剪辑", "切几个机位", "把这几个视频合成一个", "combine these angles", "PiP overlay".

Thin orchestrator for the end-to-end video localization pipeline. Routes to the four focused sub-skills — /wjs-transcribing-audio, /wjs-translating-subtitles, /wjs-dubbing-video, /wjs-burning-subtitles. Use when the user asks for full localization in one go ("帮我把这个西班牙语视频做成中文字幕+配音", "translate and dub this video", "做完整的本地化"). For any individual step (just transcribe, just translate, just dub, just burn), invoke the sub-skill directly — it's faster and the boundary is cleaner.

Use when 王建硕 wants to turn his uploaded VoiceDrop voice memos into 微信公众号 article drafts — pulling the unprocessed recordings sitting on jianshuo.dev/files (the R2 inbox), transcribing them, and mining articles from each. Triggers — "处理 VoiceDrop 录音", "把新录音挖成文章", "口述备忘变文章", "处理一下我的录音", "/wjs-mining-voicedrop".

Use when the user has a video's SRT subtitle file — a 王建硕 monologue / 讲解, OR a 对谈 / 访谈 where 王建硕 is one of the speakers — and wants to mine it into multiple standalone 微信公众号 articles, one article per distinct topic. Triggers — "把这个视频写成文章", "从字幕里挖文章", "这个 SRT 能写几篇", "把对谈写成文章", "/wjs-mining-articles <srt>".

Use when the user has one or more video clips and wants to add post-production on top — AI-generated cover as first frame, HTML/CSS captions synced to SRT, kinetic illustration overlays at hook moments, chapter chips, end-card CTA, or any other timed motion graphics. Most often used as the downstream of `/wjs-segmenting-video` — pick up where that skill stopped (raw cropped clip + per-clip SRT) and produce the upload-ready MP4. Backed by HyperFrames so everything compiles to ONE final encode — no cascade of re-encodes. Triggers — "加封面", "加字幕", "加动画", "加 CTA", "做后期", "post-production", "title card", "kinetic captions", "end card".

Use when the user wants to convert a video between horizontal and vertical orientations while preserving the inverted aspect ratio (16:9 ↔ 9:16, 4:3 ↔ 3:4, 21:9 ↔ 9:21). The skill crops a narrow band from the source and tracks the active speaker — the person whose mouth is moving — via MediaPipe face landmarks and mouth-aspect-ratio variance, so the talker stays in frame even when other people are visible. Triggers — "横转竖", "竖转横", "做成竖屏发抖音/视频号/小红书", "16:9 to 9:16", "make this vertical for Reels / TikTok / YouTube Shorts", "crop to portrait", "convert to landscape".

Use when the user has a long-form video (interview / lecture / podcast / conversation) and a transcript SRT, and wants to extract 3–6 stand-alone topical short clips from it. This skill ONLY cuts and crops — it produces raw clips + per-clip SRTs as a hand-off package for downstream post-production (`/wjs-overlaying-video`). Triggers — "切成几段", "分主题", "拆成短视频", "切片", "topic segments", "split into clips".

Use when the user has 2+ video / audio recordings of the same event captured by different devices (cameras, phones, separate audio recorders) and wants them aligned to a single common timeline. Outputs only a lightweight `.sync.json` sidecar per input — original files are never re-encoded. Triggers — "多机位同步", "对齐这几个机位", "match camera timelines", "sync these angles", "audio drift between cameras", "separate audio recorder", "Riverside / Zoom recording that needs to line up".

Use when the user wants to teach / learn an English word as a video — turn a single English word into a self-contained HLS "supercut" lesson built from the mira video base. Stitches every season2 clip where the word is spoken (via the search-app API) into one .m3u8, prepended with a Claude-written bilingual word-intro card (word + IPA + 中文 gloss + usage, Volcano TTS) and appended with a 关注王建硕 CTA card. No MP4 burn. Triggers — "teach <word>", "讲讲 <word>", "学英语 <word>", "把 <word> 做成视频", "/wjs-teaching-english <word>".

Use when the user has audio or video and wants a timestamped transcript (SRT) in the source language. Routes by source language — Chinese defaults to Volcano (豆包) ASR; other languages (Spanish, English, Portuguese, French, Italian, Japanese, Korean, etc.) use OpenAI Whisper API with word-level timestamps and self-assembled cues. Outputs SRT with punctuation-bounded cues capped for on-screen reading. Triggers — "转写", "转成字幕", "做 SRT", "transcribe", "make subtitles", "speech to text", "出字幕".

Use when the user has an SRT (or transcript text) in one language and wants it translated to another, with punctuation-bounded re-segmentation so cues end at real sentence breaks. Simplified Chinese (zh-CN) and English (en) are first-class targets; other targets follow the same rules. Outputs a target-language SRT or bilingual SRT — no audio, no burn-in. Triggers — "翻译字幕", "翻成中文", "translate this SRT", "中英双语字幕", "把这个 SRT 翻译成 X", "bilingual subtitles".

Upload one or many videos to YouTube. Use when the user wants to "上传到 YouTube", "发 YouTube", "批量上传", "upload to YouTube", "post videos to YouTube", or to publish a finished `final/` directory of MP4s. Reads per-video metadata (title / description / tags) from a sibling `UPLOAD_META.md` file when present (the user's standard markdown format), or from command-line flags. Survives behind a SOCKS/HTTP proxy by using `requests` directly for the resumable upload (the stock `google-api-python-client` MediaFileUpload stalls under this user's proxy setup).

Use when 王建硕 wants to decide whether a VoiceDrop article should get an AI 题图 (公众号封面), and which style of image prompt to use — before generating any cover. Triggers — "这篇要不要配题图", "选个题图风格", "给这篇出个题图 prompt", "choosing cover", "/wjs-voicedrop-choosing-cover".

Turn a research brief or finished script into a reviewed, narrated, rendered, and verified vertical explainer video. Use when an agent must create a short 9:16 explainer, run a no-key local demo, enforce a pre-render review gate, diagnose the video toolchain, or verify a generated MP4 before any manual publication step.

| 触发词：「音乐制作」「混音」「母带」「编曲制作」「music production」

| Pieter Levels (@levelsio) 视角. 海外 indie hacker / solo builder 极致代表 — Nomad List + Photo AI + Interior AI + Remote OK 多产品 portfolio, 单人 $250K+/月 MRR, 零员工. 把"用 AI agent 赚钱"从"融资 + 招人 + scale 到 unicorn"的 VC 操作系统, 翻成"audience first + ship daily + stay solo + $10K MRR = 自由"的 indie 操作系统. 用 Pieter 视角先反问"你要的到底是 freedom 还是 valuation" — 如果是前者, 那 path 跟 VC 派完全相反. 当用户提到 "Pieter Levels" / "@levelsio" / "indie hacker" / "solo founder" / "Nomad List" / "Photo AI" / "build in public" / "ship daily" / "$10K MRR" 时使用. 即使用户只是说 "我想做个 SaaS 副业" / "怎么 bootstrap" 也应触发.

| Trigger this skill when the user works on WeChat private-domain operations — 微信生态(企微 + 个人号 + 公众号 + 视频号 + 小程序)的私域操盘 problems and wants industry-grade thinking, tool selection, or workflow guidance. 触发词：「私域」「私域运营」「私域操盘」「微信私域」「企业微信」

| Trigger this skill when the user works on short-video paid advertising (China platforms) problems and wants industry-grade thinking, tool selection, or workflow guidance. 触发词：「投流」「巨量引擎」「千川」「信息流」「创意素材」

| 触发词：「语音转文字」「语音识别」「asr」「speech to text」「stt」

| Chris Sherwood / Crosstalk Solutions 视角 — UBNT-native 实战派的代表 YouTuber (475K+ 订阅, 1000+ 视频, UI Toolkit 作者). 加载后让 AI 用 "先演示再讲理论" / "UBNT GUI 默认不够好" / "生态绑定换性价比" 的视角看 UniFi 部署 — 尤其适合 SMB / 家庭实验室 / MSP 场景. 即使用户只是说「这个 UDM-Pro 该怎么开箱配」「80 人办公 UBNT 怎么选硬件」「我家 UniFi 慢」也应触发.

Specialized visual and multimedia processing tools. Use this skill whenever a task involves complex visual content — UI mockups, dense screenshots, design images, charts, artwork — where precise details like spacing, hex colors, font sizes, and component hierarchy need to be extracted accurately. Also use for: reviewing or auditing existing UI against designs, comparing screenshots for visual regressions, transcribing audio/video, extracting data from PDFs with complex layouts, and generating images. Trigger whenever the user wants to implement from a design, review or compare UI screenshots, analyze visual details precisely, describe artwork or aesthetic content, or process any media file (audio, video, PDF).

Measure AI Search share of voice for a target domain versus competitors across ChatGPT, Perplexity, Gemini, Google AI Overview, and AI Mode. Pulls the AIO leaderboard, then samples prompts where each domain appears as a source or brand mention, and analyses topic clusters each brand owns. Use when the user asks for AI Search share of voice, LLM visibility tracking, AEO/GEO analysis, AI Overview competitive analysis, or wants to know which brands LLMs cite in their category.

Apply Flowkit CSS naming system in Webflow. Use when creating classes, auditing existing naming, or building new components following Flowkit conventions. Flowkit is Webflow's official CSS framework with utility-first approach. Runs headlessly against a page ID — no Designer connection required, except for the optional live-canvas conveniences (current page, interactive element selection).

Compress and convert CMS item image fields to webp or avif in a Webflow collection. Prompts for collection ID, item ID, image fields, quality, and target format, then downloads, converts, re-uploads via presigned S3, and publishes the updated item.

Analyze discourse structure using InfraNodus optimize_text_structure tool and shift perspective based on diversity score. Use when a conversation, text, URL, or YouTube video would benefit from additional viewpoints, when the user feels stuck in one frame of thinking, when research needs broadening, when analyzing any text for structural balance, or when the user asks to "shift perspective", "broaden the view", "what am I missing", "analyze the structure", "check for bias", "develop this further", or "give me a different angle". Also trigger when the user provides a text or URL and asks for critical analysis, development suggestions, or wants to understand what perspectives are underrepresented. Works hand-in-hand with cognitive-variability skill for state-based interventions after structural diagnosis.

>- Analyze arguments, debate tactics, and rhetorical moves across three analyze a debate, comment thread, speech, article, or any argumentative text — identifying moves, scoring effectiveness, exposing hidden assumptions, tracking logical gaps, and checking for asymmetric standards. When InfraNodus MCP tools are available, ALWAYS begin with generate_topical_clusters, generate_content_gaps, and optimize_text_structure before linear reading. Also use to stress-test arguments or understand why something feels persuasive but wrong. Trigger a good point", "what rhetorical moves are being used", "why is this persuasive", "break down this debate", or when a user shares text asking what's going on rhetorically.

> Text network analysis, knowledge graphs, content gap detection, SEO/GEO optimization, structured memory, and text comparison via the InfraNodus MCP server (mcporter). generate research questions or ideas, compare texts, optimize text/content for SEO, analyze Google search results/queries, retrieve from a knowledge base (GraphRAG), save/retrieve structured memories, develop latent topics, or bridge conceptual gaps. Supports plain text, URLs (including YouTube video transcription), and existing InfraNodus graphs.

Refine texts in any language: perfect grammar and spelling, paraphrase ideas, avoid AI detection while maintaining authentic voice. Detects grammatical patterns that signal cognitive states or structural issues—acting as a sensory system for deeper strategic insights.

Generate high-CTR YouTube thumbnails, titles, and video scripts optimized for virality. Use when the user asks to create YouTube video titles, thumbnail text/concepts, video scripts, optimize existing titles for CTR, analyze why a video isn't performing, or needs help making content more clickable. Covers thumbnail wording, title copywriting, video script structuring (Pattern Interrupt → Hook → Framing → Curiosity Loop → Escalation → Payoff → Relevance Bridge → Loop Reopen), and YouTube algorithm optimization.

Use when a Xiaohongshu or WeChat post needs a truthful finished cover, a cover-template recommendation, or an existing-cover diagnosis and redesign that must preserve factual claims, visual anchoring, and thumbnail readability.

Use when a Xiaohongshu post, article, work report, teaching material, or chart needs a material-style center explanation image, mechanism diagram, flow diagram, chart beautification, or reusable visual material with readable Chinese labels.