agentos-skills-public

Audit and tune ANY existing site (static HTML, Astro, Next.js, WordPress export...) for search + AI discovery: SEO (Yandex & Google), AEO (AI answer engines), GEO (LLM/agent discovery). The agent checks robots.txt, sitemap.xml, canonical, meta/OG tags, JSON-LD, llms.txt and IndexNow, then generates whatever is missing. Use when the user says: optimize my site for search / SEO / чтобы находили в поиске / хочу в Яндекс и Google / AI-ответы. Tunes an existing site — it does not build one.

Generate Instagram carousels: 1080x1350 PNG slides from text (headline + body) and user photos, laid out in HTML/CSS and rendered headless via Playwright. Full design system baked in (Montserrat; cover/content/CTA card roles; calculated photo height; text-safe zones over faces; hyphen-only punctuation). Coral default palette, fully restylable to any brand (e.g. EdgeLab lime). Use WHENEVER you need a carousel for an Instagram post, want to lay out post slides, or turn a text + photos into a swipeable post. Triggers: "make a carousel", "carousel", "carousel post", "post slides", "slides for a post", "карусель", "слайды для поста", "пост-карусель", carousel/carousel. NOT for: single covers, reels, stories.

Generate premium, high-CTR YouTube thumbnails (1920x1080) — your real face from video frames, a punchy headline, a brand logo, on a dark high-contrast scene. Use this WHENEVER you need a YouTube cover/обложка/thumbnail, want a face pulled from a video onto a cover, want to "play with the font" on a cover, or want a thumbnail in a specific brand palette. Three engines: (A) GPT full-generation via gpt-image-2 with face+logo reference photos; (B) Compose — real rembg face cutout + scene + text via Playwright; (C) Hybrid — GPT no-text base + compose text for exact font control.

Research YouTube for content strategy — find what's taking off in a niche, scan competitor channels, rank videos by views and by view-velocity, and pull any video's transcript. Use this WHENEVER you need to research YouTube, find trending/viral videos on a topic, see what's blowing up on YT, scan a competitor's channel, get a YouTube video's transcript or 'what's in this video', or gather YouTube content ideas from what's working — even if the skill isn't named. Search and stats run keyless via yt-dlp; transcripts go through TranscriptAPI; a YouTube Data API key is optional for cleaner search.

Unified Telegram core (HTTP API + MCP stdio + tgdl + sales hooks) with single-session Telethon architecture. Use when setting up Telegram automation, reading/sending Telegram messages via local API, exporting chat history, or troubleshooting Telethon auth/session conflicts.

Self-audit, code quality improvement through clarifying questions, and proactive learning. Use when writing code, making decisions, completing tasks, or during weekly self-audit. Triggers on code review, task completion, weekly introspection, and when user says "интроспекция" or "аудит".

How to author, ship, and update content (lessons, bonuses) for the AgentOS intensive platform (<your-agentos-domain>). Covers source-of-truth pattern, mock layer, Supabase DB layer, block types, frontend rendering rules, deploy pipeline (build → tarball → prod server → PM2), and known pitfalls. Use when adding a new bonus, rewriting an existing lesson, changing block types, or shipping content updates to prod. Triggers: "добавь урок", "новый бонус", "обнови контент", "update lesson", "add bonus", "lesson 10 / 11 / 12", "intensive-agentos-platform", "<your-agentos-domain> content", "patch_db", "lessonN_steps". NOT for: backend logic changes, auth flows, payment integration, infrastructure work outside the content pipeline.

Archive messages from a Telegram business chat. Saves each message to daily markdown, classifies (DECISION/TASK/OPEN_QUESTION), makes daily summary, pushes decisions and summaries to storage. Use when: (1) a message is received from the monitored chat, (2) @mention of the monitor agent in this chat, (3) request for flush/day summary, (4) request for data from chat history of a partnership. NOT for: messages from other chats, direct user commands outside partnership context.

Content engine. Classify incoming content, write posts, market reports, YouTube scripts, X posts, market reviews, voice/audio routing, cross-post, screenshot handling. Triggers: "write a post", "make a post", "post about", "record idea", "save as reference", "save", "this is my stance", "my position", sending a link, repost, tweet, screenshot, audio or voice message.

Cross-review PRs in team repos. Use when reviewing PRs from other agents, creating PRs that need cross-review, or checking change levels and merge rules.

Create Datawrapper charts/tables from CSV, JSON, or inline data; publish and export PNG. Use when user wants a chart, embeddable visualization, or PNG for Telegram.

Dev Pipeline v4 -- полный цикл разработки. Codex аудит/план → Coder субагент пишет код → Claude+Codex cross-review → Coder фиксит → результат user. Вызывай когда user даёт задачу на код.

Генерация Excalidraw-схем (.excalidraw v2) из JSON. Поддерживает pipeline, mindmap, flowchart, sequence, freeform. Цветовая палитра и формат по стандартам MCP Excalidraw. Экспорт файла для excalidraw.com.

Lightweight 4-phase pipeline for small, well-scoped coding tasks (50-300 LOC, 1-3 files, single subsystem, 15-30 min). Plan (Codex) -> Implement (Opus) -> Dual Review (Codex+Opus) -> Ship. NO parallel research, NO audit phase, max 1 fix iteration. Plan single-model (Codex GPT-5.5) to offload Opus subscription; Review still dual because review is where bugs are caught cheap. Trigger on: short feature, small refactor, single hook, single script, isolated bug-fix-with-test, focused config rewrite, adding CLI subcommand. Russian: быстрый рефакторинг, мелкая фича, новый скрипт, хук, скрипт, исправление с тестом, фастлуп, fast-loop-coding, быстрый луп. Do NOT use for: 5-line fixes (just edit), migrations / rewrites / multi-file features (use loop-coding), production deploys, security-sensitive changes, RED files (CLAUDE.md / rules.md / USER.md), tasks crossing 3+ services. ESCAPE: if scope grows past 300 LOC or 3 files mid-task, abort and switch to loop-coding.

Transcribe voice messages (.ogg) via Groq Whisper API. When you see <media:audio> in a message, extract the file path and transcribe it using the included script.

Generate images/videos via Higgsfield AI. Models: GPT Image 2, Nano Banana 2/Pro, Soul V2/Cinema/Cast/Location, Seedance 2.0, Veo 3.1, Kling 3.0, Flux 2, Z Image, Hailuo; plus Marketing Studio ads with avatars/products/hooks/settings. Use when: "generate an image", "make a picture", "make a video", "animate this photo", "image-to-video", "edit/stylize/remix this image", "produce a clip", "create an ad", "make a UGC video", "product demo", "unboxing", "TV spot", "brand video", "presenter video", "import product from URL", "create avatar for ad". Supports text-to-image, image-to-image, image-to-video, reference-based generation, and Marketing Studio. Auto-detects upload IDs vs previous job IDs. Chain with higgsfield-soul-id when the user wants face/identity consistency. NOT for: training Soul Character (use higgsfield-soul-id), product photoshoots (use higgsfield-product-photoshoot), marketplace listing cards (use higgsfield-marketplace-cards), text/chat/TTS tasks.

Full toolkit for Instagram: account analytics via HikerAPI SaaS + downloading reels/posts via Cobalt API. Use when: (1) download a reel/video/photo from Instagram, (2) find top-performing reels of an account, (3) analyze a competitor (views, likes, comments, engagement), (4) manage a watchlist of accounts for monitoring, (5) download + transcribe video, (6) hashtag/location search. Triggers: "download reel", "download from instagram", "analyze account", "viral reels", "top reels", "watchlist", "track account", "competitor instagram", "engagement", links to instagram.com/reel/, instagram.com/p/, instagram.com/stories/. NOT for: YouTube, Twitter, mass spam, auto-likes, auto-follows.

Learnings System v2 — self-improvement через scoring ошибок. 3 слоя: Episodes (сырой лог) → Learnings (scored) → Rules (promoted). Используй когда: (1) пользователь поправил действие, (2) обнаружена ошибка, (3) нужен отчёт по learnings, (4) lint/audit накопленных уроков.

Orchestrates implementation-heavy coding work — migrations, major refactors, rewrites, multi-file feature builds, auth/API overhauls, system rebuilds — via a 7-phase pipeline (research, audit, plan, implement, review, fix-loop, ship). Scripts-first (70%+ automated), dual-model critique (Opus + Codex), auto-generated helper sub-skills via skill-creator, staging-first deploy. Trigger on: migrate, rewrite, rebuild, overhaul, large feature — including Russian миграция, крупный рефакторинг, переписать, переделать систему, большая фича, лупкодинг. Do NOT use for: single-file edits, fixes under 5 lines, dependency bumps, typo or link fixes, docs-only changes, adding a log line or type hint, code review of existing PRs, ADR or planning-only requests, content writing or naming tasks (use gsd-plan-phase or brainstorming for planning).

Clean Markdown extraction from any URL via markdown.new service. 80% token reduction for LLM use. Handles PDFs, images, audio, and embedded media. No signup required. Use when web_fetch fails or returns too much HTML noise. Triggers: "markdown.new", "clean markdown", "extract content", "convert to markdown"

Design and build production REST APIs and MCP servers following industry standards (Google AIP, Microsoft Azure, Zalando, Stripe, Anthropic MCP Spec 2025-06-18). Use when: (1) designing new API endpoints from scratch, (2) building MCP server for an existing API, (3) converting REST endpoints to MCP tools/resources, (4) reviewing API architecture, (5) writing OpenAPI spec, (6) auditing API security and production readiness, (7) debugging API anti-patterns. Triggers: «build API», «design endpoints», «create MCP server», «MCP», «API architecture», «OpenAPI spec», «REST to MCP», «спроектируй API», «сделай MCP сервер», «антипаттерны API», «review API design».

Самоаудит памяти и состояния агента. Используй когда: (1) user пишет «аудит памяти», «проверь память», «memory-audit», (2) после крупных изменений. Проверяет: memory backend, личную память, скиллы, cron, heartbeat.

Create lesson presentations and slide layouts on Miro board via MCP API. Use this skill when: (1) user asks to "build a presentation for a lesson", "make workshop/course slides", "slides on Miro", (2) need to create 6+ frames 1920x1080 for video recording, (3) course lessons in Miro, (4) Convert to Slides in Miro, (5) text on the left under face-cam on the right. The skill includes a ready-made system: dark graphite + gold style, font scale, Miro DSL coordinate rules, bulk_create limits, recipe for assembling 6-frame presentations. Triggers: "slides for lesson", "lesson presentation", "miro board", "miro-board", "build slides", "make presentation", "workshop slides", "course slides".

Веб-ресёрч через Perplexity Sonar API. Используй когда нужен актуальный поиск в интернете, факт-чек, анализ трендов, best practices. Триггеры: «загугли», «найди», «ресёрч», «что говорят про», «best practices», «как правильно», «perplexity».

Generate beautifully formatted HTML presentations with a dark theme. Triggers: /present, 'create presentation', 'visualize', 'make HTML'. Accepts any content (lesson, report, data, plan, comparison, roadmap) and produces a standalone HTML file.

Create a production brief (ТЗ) for Instagram Reels: teleprompter script and editor brief. Triggers: /reels, 'write a reel', 'reel brief', 'create reel script', 'editor brief', link to an Instagram reel + adaptation request. Downloads a reference reel, transcribes it, analyzes structure, generates an HTML doc with two blocks: teleprompter script (with timings) and editor brief (visuals, transitions, on-screen text). Format: HTML with dark theme. NOT for: posting reels, video editing, account analytics.

Use when auditing or repairing Linux or macOS hosts that run OpenClaw, Telegram bots, or related automation, especially when the operator needs to map hosts, runtimes, and safe access paths before diagnosis.

Converts a Telegram post (typically from a channel) into a Threads thread of 5-8 posts with a banner for the first message and a CTA with UTM at the end. Includes 4 banner templates (negative/positive/transformation/equation in 3D app-icon style + black background) and a proven 7-block thread structure. Use ALWAYS when user says "make a thread", "adapt for Threads", "turn this post into a thread", "take this post and break it into a thread", "find a popular post and make a thread", or shares a forwarded Telegram post and asks for Threads content.

Monitor topics of interest and proactively alert when important developments occur. Use when user wants automated monitoring of specific subjects (e.g., product releases, price changes, news topics, technology updates). Supports scheduled web searches, AI-powered importance scoring, smart alerts vs weekly digests, and memory-aware contextual summaries.

Транскрипт YouTube роликов через TranscriptAPI.com. ОБЯЗАТЕЛЬНО используй этот скилл когда: (1) пользователь присылает YouTube ссылку (youtube.com, youtu.be), (2) просит рассказать о чём видео, (3) просит транскрипт/субтитры, (4) говорит «что в этом видео», «о чём ролик», «перескажи видео», (5) любой запрос связанный с содержимым YouTube видео. Триггеры: YouTube ссылка, «транскрипт», «о чём видео», «что говорят в видео», «перескажи ролик», «summarize video», «get transcript». НЕ использовать yt-dlp или youtube-transcript-api -- только этот скилл.

Read Twitter/X tweets, articles, threads, profiles, and search. Use when the user shares a Twitter/X link, asks to read a tweet/article, check a profile, search tweets, or fetch tweet content. Covers both free (FxTwitter) and paid (SocialData API) methods with automatic fallback.

YouTube producer for a creator's channel. Generates 5 title variants + thumbnail concept based on an idea from a tweet/article/video, using proven formulas from top YouTube creators (e.g. Nate Herk: 680K subs, 116K avg views). Also provides a hook for the first 30 seconds and a video structure using his patterns. ALWAYS use this skill when: (1) user shares a tweet, article, or another video and says "make a video on this topic", (2) asks for "video title", "come up with a video name", "YouTube title", (3) asks for "thumbnail variants", "cover concept", "thumbnail for a video", (4) asks for "video hook", "how to start the video", "first seconds of video", (5) preparing YouTube content — even if user doesn't name this skill explicitly. Triggers: YouTube, video, title for video, thumbnail, cover, hook.