qu-skills/skills
26 skills634k installs13.3k starsGitHub
Install
npx skills add https://github.com/qu-skills/skillsSkills in this repo
1Ai Video GenerationAI Video Generation is an agent skill that wraps the inference.sh belt CLI so you can produce AI videos from natural-language prompts and reference media. It catalogs dozens of hosted models—including Google Veo, Seedance, HappyHorse, Wan, and Grok—and maps them to jobs like text-to-video, image-to-video, lipsync, and upscaling. Solo and indie builders use it when they need launch-ready motion content without hiring an editor or wiring each provider API by hand. You log in with belt, pick an app ID for the model, and pass JSON inputs from your coding agent. The skill fits the Prism journey across Validate (landing hero clips), Launch (distribution creatives), and Grow (ongoing social content), while staying a task integration rather than a methodology. Pair it with remotion-render when you need code-driven motion graphics instead of pure generative video.100kinstalls2Remotion RenderRemotion Render is an agent skill that sends React/Remotion component code to inference.sh’s remotion-render app and returns a finished video file. Instead of dragging timelines in a GUI, you express motion with familiar frontend primitives—frames, sequences, springs, and interpolations—and let belt handle cloud render. Solo builders adopt it when dashboards, changelogs, or metrics need repeatable branded intros/outros generated from the same codebase as the product. The skill is a generator-style integration: log in with belt, pass escaped TSX in JSON, and tune fps and dimensions. It pairs naturally with AI Video Generation when you want generative B-roll layered under code-driven typography. Placement spans Build (author components), Launch (export launch reels), and Grow (automate weekly stats videos), while the Prism shelf stays Build/frontend as the home for TSX work.100kinstalls3Ai Image GenerationAI Image Generation is an agent skill that wires your coding assistant to inference.sh’s belt CLI so you can call dozens of image models from a single Bash workflow. Solo and indie builders use it when they need hero images, app screenshots, ad creatives, or concept art without opening a separate Midjourney or DALL·E tab. The skill documents model picks (FLUX LoRA variants, Gemini, Grok, Seedream, Reve, GPT-Image-2), capabilities like inpainting and upscaling, and quick-start commands after belt login. It fits the Prism journey wherever visuals unblock shipping: polishing UI during Build, assets for a Validate landing page, or Launch and Grow social posts. Complexity is intermediate because you manage API access, model IDs, and JSON payloads, but the pattern is repeatable once belt is installed.100kinstalls4Twitter AutomationTwitter Automation is an agent skill that exposes Twitter/X operations through the inference.sh belt CLI as discrete apps for posting, engagement, and lightweight CRM-style actions. Indie founders use it to announce launches, reply with consistent tone, and run small growth experiments without maintaining a custom X API client in their repo. Each capability maps to a belt app ID—post-tweet for text, post-create for media, plus like, retweet, DM, and follow helpers documented in the skill tables. It is an integration skill with network and credential implications: you authenticate via belt login and pass structured JSON from your coding agent. Prism places it on Launch/distribution as the primary shelf, with Grow/lifecycle as a secondary home for ongoing engagement. It complements content skills that draft copy but do not publish.100kinstalls5Ai Avatar VideoAI Avatar Video connects your agent to inference.sh’s belt CLI to create lipsynced talking heads and virtual presenters from a portrait plus script or audio. Indie builders use it for explainer clips, UGC-style ads, dubbing, and lightweight spokesperson videos without managing a separate video SaaS UI. The skill highlights P-Video-Avatar as the default economy path, optional heavier models, and TTS backends including Inworld with broad language support. Placement in Prism emphasizes Launch distribution and Grow content, though Validate prototype demos can use short presenter clips. You need belt installed, network access, and comfort passing image URLs and voice parameters as JSON. It is an integration skill, not a full video editor—best for short-form presenter output your agent can script and regenerate.100kinstalls6Agent BrowserAgent-browser documentation in the qu-skills collection teaches procedural patterns for driving a real browser from agent tooling—starting sessions, interacting with refs, and verifying state after navigation. The authentication-focused module walks through standard login, OAuth and SSO, two-factor challenges, session reuse, and cookie handling so solo builders do not reinvent fragile scripts for every integration test or internal admin workflow. Examples assume the belt `agent-browser` app with `open`, `interact`, and `snapshot` functions and jq-parsed session identifiers, which maps cleanly to agents that need repeatable, scriptable UI access rather than one-off manual QA. Use it while wiring agent features that depend on logged-in SaaS dashboards, customer portals, or OAuth consent screens. It complements session-management references in the same skill family for persistence details. This is integration knowledge, not a replacement for human security review of credentials and tenant data policies.18.7kinstalls7Web Searchweb-search is a journey-wide agent skill that connects solo builders to Tavily and Exa capabilities through the inference.sh belt CLI. Instead of hand-rolling HTTP clients, you invoke documented apps such as tavily/search-assistant for AI-backed answers, tavily/extract for page content, and exa/search, exa/answer, and exa/extract for complementary retrieval flows. Triggers span market research before you commit to a product, fact-checking during content or docs work, and feeding RAG or agent pipelines during Build. Installation is explicitly coupled to belt: login once, then run structured JSON inputs from Bash-capable agents. The skill is an integration layer, not a hosted search API key wizard—you still need belt accounts and network access. Intermediate comfort with CLI flags and JSON helps; beginners can copy the Quick Start verbatim. Use when Perplexity-style answers must run inside Claude Code or Cursor automation. It does not replace your own citation policy or rate-limit monitoring. Pair with content skills after research when turning findings into landing copy or ADRs.12.5kinstalls8Agent ToolsThe agent-tools skill teaches Claude Code, Cursor, Codex, and similar agents how to discover inference.sh Belt apps through the belt CLI. Solo and indie builders use it when they are extending an agent with generative media, TTS, or other hosted capabilities and need a repeatable way to search the public store, filter by category, paginate results, and pull JSON schemas before calling belt app run. The readme documents your-apps commands, store browsing, featured and newest listings, detailed tables, and popular slugs such as falai/flux-dev-lora and google/veo-3-1-fast grouped by image and video generation. It pairs naturally with belt-sh CLI install skills and downstream executor or model skills once you have chosen an app ID. Treat it as catalog navigation procedural knowledge—not a sandbox runtime itself.12.4kinstalls9Python ExecutorThe python-executor skill packages procedural knowledge for executing Python inside inference.sh’s safe, CPU-only sandbox using the belt CLI. It targets solo builders who want their coding agent to run short scripts for data wrangling, scraping, visualization, media processing, or glue automation without shipping a custom Docker image. The documented app ID is infsh/python-executor with an input schema centered on a code string plus environment limits for memory and timeout. Install guidance points to npx skills add belt-sh/cli and belt login before belt app run. Because the environment ships a broad scientific and automation stack, the skill spans quick validation scripts and build-time integration checks, but it is not a long-running production worker—use it for bounded agent tasks. After discovery of related apps, pair with agent-tools or infsh-cli skills when you need a different hosted capability than raw Python.12.4kinstalls10Infsh CliThe infsh-cli skill is procedural documentation for discovering and inspecting apps on inference.sh using the belt command-line interface. Solo builders install it when their agent must navigate a large public store of image, video, audio, and text models without hard-coding slugs. Coverage mirrors core discovery flows: personal app lists, paginated store views, category and keyword search—including both belt app store search and the --search flag variant—featured and newest sorts, JSON export via --save, and detailed retrieval with belt app get. Popular examples such as FLUX LoRA variants, SDXL, Gemini image preview, Grok imagine, and Veo video models anchor category browsing. It complements belt-sh CLI installation skills and downstream execution skills once an app is selected. Position it as CLI reference knowledge for agent sessions rather than application business logic.12.4kinstalls11Landing Page DesignLanding Page Design is an agent skill for solo builders who need marketing pages that convert, not generic wireframes. It walks through above-the-fold structure, hero section design, CTA psychology, social proof placement, F-pattern reading, and mobile layout rules—then ties those decisions to actionable belt commands on inference.sh for hero visuals and competitor landing-page research. Use it when you are scoping a validate-phase landing page, refreshing a SaaS product page before launch, or optimizing an existing page’s conversion rate. The workflow assumes you install and authenticate the belt CLI, then run documented apps for image generation and search. It complements front-end implementation skills by defining what to build and why, while generating assets that match a stated visual brief rather than leaving heroes as placeholder blocks.6.3kinstalls12Storyboard CreationStoryboard Creation helps solo builders and creators turn scripts into visual pre-production using standard film shot vocabulary, continuity rules, and annotated panel layouts. Install it when you are storyboarding video ads, music videos, animation beats, or any project that needs a shot list and compositional plan before cameras roll. The skill documents abbreviations for shot types and angles, when to use each framing, and how to keep continuity across panels. Operational flow assumes the inference.sh belt CLI: log in, generate 16:9 cinematic panels with flux-dev-lora, then stitch images into a horizontal board. That makes it a generative-media workflow skill—not a generic drawing prompt—tied to explicit Bash(belt *) tooling. It suits indie filmmakers and marketers who want agent-assisted panels without abandoning professional storyboard conventions. Use it in validation to prototype the visual script; reuse in growth when planning new creative cuts for distribution.6.3kinstalls13Youtube Thumbnail DesignYouTube Thumbnail Design is an agent skill that turns platform specs and CTR psychology into repeatable thumbnail workflows for solo builders publishing video. It defines dimensions, aspect ratio, file limits, and mobile-safe zones so titles and faces are not cropped in the sidebar feed. The skill pairs editorial rules—high contrast, expressive faces, shallow depth of field, limited on-image text—with AI generation through the inference.sh belt CLI, including example Flux LoRA runs at 1280×720. Use it when you need video cover images, click-through optimization, or consistent visual packaging across a channel rather than one-off graphic design in Figma. It fits creators and indie SaaS marketers who ship tutorials or launch content without hiring thumbnail artists, and it complements broader distribution work in Grow by making preview frames testable and on-brand.6.3kinstalls14Character Design SheetCharacter Design Sheet is an agent skill for solo builders who need the same character to look identical across dozens of AI-generated panels, sprites, or marketing stills. It explains why naive prompting fails, then walks through reference sheets—front turnaround, expressions, palette locks—and LoRA-oriented generation via the inference.sh belt CLI with concrete Flux examples on white-background concept art. Triggers span character consistency, OC design, and animation prep. The skill suits indie game artists, comic creators, and narrative SaaS demos that rely on recurring mascots without a full art department. It complements generative media workflows in Build and early Validate prototype visuals when you are proving a style before production volume. Expect procedural knowledge and prompt patterns, not automated rigging or 3D mesh export.6.3kinstalls15Product PhotographyProduct Photography is an agent skill for solo builders and small brands who need listing-ready visuals without hiring a studio. It documents how to invoke inference.sh through the belt CLI to produce packshots on white backgrounds, aspirational hero angles, and lifestyle contexts using structured prompts around lighting, shadows, and sharp commercial focus. The workflow fits when you are polishing a Shopify or Amazon catalog, refreshing creatives for paid social, or prototyping packaging shots before a photoshoot. It is a procedural prompt-and-CLI guide rather than an in-editor image plugin, so your agent composes the JSON input and runs `belt app run` with size and ratio parameters you choose. Expect to iterate prompts for brand consistency and to validate output against each channel’s image rules yourself. Pair it with video-ad or landing skills when you need motion or full-page context around the same product story.6.3kinstalls16App Store ScreenshotsApp Store Screenshots is an agent skill for solo and indie mobile builders who need store-ready visuals without guessing platform rules. It documents exact iOS and Android screenshot and preview specs, gallery structure, and mockup conventions, then wires those requirements to inference.sh belt commands that generate polished device scenes from text prompts. Install it when you are polishing an App Store or Play listing, refreshing ASO creatives, or producing preview-style marketing stills before submission. The skill assumes you already use belt for generative apps; it does not replace native capture from simulators but complements AI mockups and spec checklists so agents do not ship wrong resolutions or missing required sizes. For builders shipping SaaS companions or standalone mobile apps, it keeps listing work in the launch phase instead of ad-hoc image edits in chat.6.3kinstalls17Video Ad SpecsVideo Ad Specs teaches agents how to brief and generate platform-native video ads using the belt inference.sh CLI while respecting each network’s hard limits on aspect ratio, length, and format. Solo builders running their own TikTok or Meta campaigns get copy-paste commands plus specification tables so generated clips are not cropped or rejected at upload. The skill bridges creative prompting—authentic unboxing scenes, bright rooms, social-native pacing—and technical parameters such as 9:16 at 1080×1920 for TikTok or shorter bumper durations for YouTube. It also references structured messaging like AIDA so hooks and captions align with performance marketing habits. You still own compliance, rights to likeness, and final export encoding; the skill is the spec cheat sheet and generation workflow inside your coding agent. Use when launching a new offer, testing creative variants, or handing an agent a repeatable ad-production checklist.6.3kinstalls18Competitor TeardownCompetitor Teardown is a research workflow skill that turns scattered bookmark tabs into a repeatable competitive intelligence pass for solo founders. Using the belt CLI, your agent can run Tavily-backed search for category comparisons, pull live screenshots of rival sites, and organize findings through a seven-layer analysis model spanning product surface area, pricing, positioning, go-to-market, social proof, UX impressions, and defensibility signals. The output orientation is practical: matrices you can drop into Notion or slides, SWOT bullets for roadmap debates, and positioning language for landing copy or sales one-pagers. It does not replace primary customer interviews or paid data subscriptions, but it compresses the first week of desk research into agent-executable commands. Revisit the same framework during Validate when narrowing scope and during Build PM work when prioritizing differentiation.6.3kinstalls19Product Hunt LaunchProduct Hunt Launch is an agent skill for solo founders and side-project builders who want a structured Product Hunt go-live instead of improvising taglines and gallery assets the night before. It encodes listing element specs, visual dimensions suited to PH galleries, and tactical notes for launch day engagement, pairing that playbook with belt commands to generate hero imagery and search recent top SaaS launches for positioning cues. Use it when you have something shippable and need distribution on Product Hunt specifically—not as a substitute for full GTM across every channel. The skill leans on external belt apps and search assistants, so your agent needs network access and a belt account. It fits builders who already validated the idea and are in the loud, crowded launch window where creative consistency and research-backed copy matter.6.3kinstalls20Image To VideoImage to Video is an agent skill for solo builders who have a strong still frame—screenshot, product photo, or generated key art—and need controlled motion for ads, landing pages, or store previews without learning every video model by trial and error. It compares inference.sh belt apps such as Wan 2.5 i2v and alternatives, documents when each motion style fits, and supplies copy-paste command patterns linking image generation to animation with explicit prompt language for ripples, drift, and camera feel. Primary use is launch distribution content, but the same workflow applies in build when you are producing reusable media assets inside an agent-driven pipeline. Expect intermediate complexity: you must log into belt, manage file paths between steps, and iterate prompts when motion artifacts appear. It is a procedural integration skill, not a hosted video editor.6.3kinstalls21SeedanceSeedance is an integration skill for generating video through ByteDance Seedance 2.0 on inference.sh using the belt CLI. Indie creators, marketers, and app builders use it when they need text-to-video, image-to-video, or reference-to-video output with synchronized audio for social clips, product demos, music visuals, and animated explainers. The skill documents model choice between full-quality Seedance 2.0, the faster variant, and Studio mode for portrait consistency via a private asset library. Operators run authenticated `belt app run` commands with structured JSON inputs rather than hand-rolling API clients. It pairs with the belt CLI install skill and shell permissions. Not aimed at real-time in-game engines or long-form editorial video pipelines—best for short, prompt-driven generative clips during launch and content growth pushes.199installs22P ImageP-image is an agent skill for running Pruna’s P-Image family through the inference.sh belt CLI. Solo builders use it when they need fast text-to-image output, preset LoRA styling, or edit passes including multi-image compositing, without maintaining local diffusion infrastructure. The skill documents app IDs, model roles in a comparison table, and copy-paste bash entrypoints after belt login. It pairs naturally with agent workflows that need thumbnails, marketing stills, or iterative visual edits during product build. Install the belt CLI via the documented npx skills add path before invoking run commands. This is a remote inference integration skill, not an in-repo image library.198installs23Gpt ImageGPT-Image is an integration skill that wires solo builders’ agents to OpenAI GPT-Image-2 through inference.sh’s belt CLI, so image work stays scripted and repeatable instead of bouncing through a separate browser tab. After belt login, you run openai/gpt-image-2 with JSON prompts for pure generation, edits guided by reference images, or inpainting with masks—useful for landing hero shots, app store screenshots, ecommerce mocks, and quick concept art during build. Batch mode supports up to ten variants in one invocation, with common web formats on output. Triggers cover gpt-image-2, DALL·E-family wording, and ChatGPT image phrasing so discovery matches how builders actually ask. Pair with validate prototyping when you need visuals before code, or launch distribution when you refresh creatives. Not a replacement for brand design systems or vector UI kits in Figma.197installs24HappyhorseHappyHorse is an agent skill for driving Alibaba’s HappyHorse 1.0 video models through the inference.sh belt CLI. It targets solo and indie builders who need believable motion, character-consistent clips, or fast video edits without standing up their own GPU stack. You authenticate once with belt login, then call documented app IDs for text-to-video, image-to-video, reference-to-video, or video edit flows using JSON payloads. Typical jobs include animating a still for a landing hero, turning a script into a short demo reel, keeping a character aligned across shots with multiple references, or applying natural-language edits to existing footage. The skill assumes belt is installed and treats HappyHorse as a managed inference surface rather than local weights. Use it when marketing and product timelines need turnkey generative video instead of a custom pipeline.197installs25P Videop-video is an agent skill for Pruna-optimized video generation on inference.sh, exposing P-Video for prompt and image-driven clips (with audio support) and WAN models for lighter text-to-video and image-to-video at 480p/720p. Indie builders reach for it when they need b-roll, hero loops, or social snippets without running heavy local diffusion stacks or premium hosted editors. The skill centers on belt CLI installation, login, and `belt app run` examples such as drone-over-forest sunsets. It pairs naturally with launch distribution and grow content workflows, and with validate prototypes that need motion on a landing page. Complexity is intermediate due to API billing, model choice, and JSON inputs. Unlike p-video-avatar, this skill is general scene video rather than talking heads—complementary in a Pruna + belt toolchain for full funnel creative.197installs26P Video Avatarp-video-avatar is an agent skill that wraps Pruna P-Video-Avatar on inference.sh so your coding agent can generate realistic talking-head videos from a single portrait, using built-in TTS or driven audio. Solo and indie builders use it when they need AI presenters, product walkthroughs, education clips, or multilingual marketing without standing up a separate avatar SaaS stack. The skill documents Quick Start flows with `belt login` and app runs, model pairing with P-Image for portraits, and trigger vocabulary for digital humans and HeyGen/Synthesia-style workflows. It fits the launch and grow phases when distribution and content matter, and validate when you prototype landing or demo videos. You install the belt CLI skill and grant shell access to `belt` commands; outputs are renderable video assets suitable for channels, courses, and in-app explainers.197installs