skills

Use when you need to drive a real browser to open a page, fill forms, click, screenshot, scrape, or test a login flow. Triggers on "automate the browser", "test the page", "scrape this site".

Use when the user says "lightpanda", "scrape this page", "headless browse", or "dump the DOM". Fast headless browser for rendering URLs without Chromium, scraping, and CDP automation.

Use when user says "does it work?", "screenshot the page", "show me the form", "check the checkout flow", or "open the browser". Drives Chromium/Firefox/WebKit via Playwright MCP to verify storefront/admin UI changes, reproduce reported UI bugs, test Medusa dev servers on localhost, or explore third-party sites.

Track resume versions, maintain one master resume, and manage per-job tailored copies. Use when user says "manage my resumes", "track resume versions", "which resume did I send", "master resume", "organize my resumes", or "version of my resume for this job".

Use when user says "extract text from pdf", "merge pdfs", "split a pdf", "fill out a pdf form", "extract tables from pdf", or "create a pdf". Complete PDF manipulation toolkit for extracting text and tables, creating new PDFs, merging/splitting documents, and handling forms. When Claude needs to fill in a PDF form or programmatically process, generate, or analyze PDF documents at scale.

End-to-end pipeline that chains trendradar → article-writer → X thread → Higgsfield hero image → Postiz draft in a single shot, turning a current trend topic into a posted-ready social thread with a generated hero image. Use when user says "/trend-to-thread", "post a trend thread", "build a thread from trends", "új X thread mai trendből", or "Postiz X poszt mai trendekből".

Use when user says "Coolify", "deploy backend", or "check deploy logs". Env vars, builds, restarts, logs, rollback.

AI image generation Creative Director powered by Google Gemini Nano Banana models. Use this skill for ANY request involving image creation, editing, visual asset production, or creative direction. Triggers on: generate an image, create a photo, edit this picture, design a logo, make a banner, visual for my anything, and all /banana commands. Handles text-to-image, image editing, multi-turn creative sessions, batch workflows, and brand presets.

Use when user says "improve the UI", "make it tasteful", or "design pass". Hierarchy, spacing, polish, visual validation.

Use when user says "taste-skill v1", "design-taste-frontend-v1", or "legacy taste skill". The original v1 taste-skill, preserved for projects depending on its exact behavior. The current default is `design-taste-frontend` (v2 experimental), which is a substantial rewrite. Use this v1 install name only if you need exact backward compatibility.

Use when user asks to "design", "build", "review", or "improve" UI/UX with vetted recommendations. Queries Python DB of 67 styles, 96 palettes, 57 font pairings, 99 UX guidelines, 25 chart types across 13 stacks.

Performance regression detection using the browse daemon. Establishes baselines for page load times, Core Web Vitals, and resource sizes. Compares before/after on every PR. Tracks performance trends over time. Use when: "performance", "benchmark", "page speed", "lighthouse", "web vitals", "bundle size", "load time". Voice triggers (speech-to-text aliases): "speed test", "check performance".

Design finalization that generates production-quality Pretext-native HTML/CSS. (gstack)

Upgrade gstack to the latest version.

Manage project learnings.

Weekly engineering retrospective. Analyzes commit history, work patterns, and code quality metrics with persistent history and trend tracking. Team-aware: breaks down per-person contributions with praise and growth areas. Use when asked to "weekly retro", "what did we ship", or "engineering retrospective". Proactively suggest at the end of a work week or sprint.

Set up gbrain for this coding agent by installing the CLI, initializing a local PGLite or Supabase brain, registering MCP, and capturing per-remote trust policy. (gstack)

Keep gbrain current with this repo's code and refresh agent search guidance in CLAUDE.md. Wraps the gstack-gbrain-sync orchestrator with state (gstack)

Use when user says "Hostinger DNS", "DNS record", or "point domain". Records, propagation, verification, rollback risk.

Use when user says "Hostinger domain", "buy domain", or "domain settings". Registration, nameservers, ownership, renewal.

Use when user says "Hostinger hosting", "upload site", or "shared hosting". Static deploys, files, domains, SSL.

Use when user says "Hostinger VPS", "server deploy", or "VPS access". SSH, services, Docker, logs, health, safe ops.

Use when user says "audit my google ads", "review my ad spend", "find wasted spend", "check my campaigns", "where am I losing money on ads". Full Google Ads account health analysis detecting wasted spend, search term leaks, negative keyword gaps, bid strategy issues, and Quality Score problems. Use when analyzing campaign data, pasting Google Ads exports, reviewing account performance, or requesting a full diagnostic of advertising spend efficiency.

Use when user says "build a buyer persona", "who is my ideal customer", "research my ICP", "define my target audience", "who am I selling to". Build detailed B2B buyer personas with pain points, objections, buying triggers, and messaging angles. Use when given a product and market to research ideal customer profiles, creating buyer personas, or needing to understand who you're selling to before launching campaigns.

Use when user says "build programmatic seo pages", "create page templates", "scale seo content", "set up pseo", "make a comparison page template", "avoid thin content at scale". Create scalable programmatic SEO page templates with title patterns, internal linking logic, schema markup, and thin content avoidance strategies.

Create a complete brand design guide — color palette, typography pairings, UI component previews, and visual identity rules with example mockups.

Engineer professional-grade brand logos using geometric primitives and negative space — generates minimalist, scalable vector-style marks via muapi.ai

Generate a complete set of multi-angle product shots — front, side, back, top-down, and 45-degree perspective — for full product visualization.

Reasoning-driven image generation using structured creative briefs (Gemini 3 style) — generates high-fidelity images via muapi.ai with logic-based prompting

Use when user says "Medusa backend", "custom module", "API route", "workflow", or "data model". Required for Medusa modules, routes, workflows, links, business logic.

Use when user says "Higgsfield to Medusa", "generate product photos", or "import AI product assets". Asset generation → product mapping → import.

Use when user says "write a MARVA blog post", "marvahome blog", "publish MARVA blog", "generate Hungarian blog for marvahome", "MARVA Home cikk", or asks to draft, author, push, or publish an SEO blog post for marvahome.com. End-to-end: Hungarian SEO+GEO authoring → marva-blog MCP push → Higgsfield thumbnail.

Use when user says "WooCommerce import", "migrate products", or "Woo to Medusa". Extraction, mapping, assets, import.

Autonomous loop that optimizes a measurable metric (coverage, bundle size, lint errors) one change at a time. Use when repeated experiments can be judged by a mechanical score.

Generates optimized prompts for AI tools (LLMs, Cursor, Claude Code, Midjourney, image/video/voice/3D AI, workflow tools). Use when user says "write me a prompt", "build a prompt for", "fix this prompt", "improve this prompt", "adapt this prompt", "make a midjourney prompt", "claude code prompt", "/prompt-master", "sharper prompt", or pastes a prompt to refine. Skip for general conversation, code edits, or document writing.

Use when the user mentions "oh-my-agent", "OMA", a portable agent harness, vendor-agnostic skill bundles across Claude Code + Codex + Cursor, or asks how OMA compares to Colony/codex-fleet. Routes the user to the upstream tool and explains the cue-side trade-offs vs the parallel-agents stack they already run.

Use when user says "audit my predictions", "how is the model doing", "show prediction accuracy", "why is the model wrong", or "review my Brier score". Reads the local predictions.jsonl + Brier report via `polymarket-live` MCP and explains where the model agrees/disagrees with reality. NOT for opening or resolving predictions.

Use when user says "what's the Polymarket market doing", "research a Polymarket market", "BTC 5m snapshot", "look up a Polymarket market", or "what are the odds on X". Live overview via `polymarket-live` MCP — bundles BTC 5m snapshot, market search, and order book. NOT for placing orders.

Use when user says "keyword research", "SEO keywords", "search volume", "keyword difficulty", or "topic clusters". Seed terms, intent, volume, difficulty, clustering.

Use when user says "OpenAI docs", "OpenAI API", or "ChatGPT API". Official OpenAI docs lookup: current docs, examples, limits, citations.