gerdsenai

Scaffold a standalone AgenticUI web app — a frontend that shows an AI agent working: a live in-browser terminal (xterm.js + node-pty), a pseudo-desktop (Dockerized Xvfb + x11vnc + noVNC), and optional computer-use (a BYO-model VLM click/type loop). Phased, public-safe, localhost + token + read-only by default.

Build markdown files into professional PDFs — single file or entire directory

Print a curated starter prompt for /ultraplan — design-driven planning for the plugin (AgenticUI templates + scaffolder/command/README surface) with Anthropic's frontend-design skill steering aesthetic direction.

Scaffold a new cross-platform, agent-ready project based on GerdsenAI Agentic Devtools

Configure and use a local/remote OpenAI-compatible LLM backend (Ollama, vLLM, LM Studio, …)

Use this agent to scaffold, customize, or extend an AgenticUI app — a standalone web frontend that shows an AI agent working: a live in-browser terminal (xterm.js + node-pty over WebSocket), a pseudo-desktop (Dockerized Xvfb + x11vnc + noVNC), and optional computer-use (a VLM-driven click/type loop). Handles generating the app via /gerdsenai:agentic-ui, tailoring the templates, and polishing the UI. <example>Build me an agent-observability UI with a live terminal and desktop</example> <example>Generate an AgenticUI app, phase 3, read-only</example> <example>Add a logs panel to the generated AgenticUI frontend</example> <example>Make a Manus-style view where I can watch my agent work</example>

Use this agent when the user asks to create a report, write a document for PDF output, generate formatted documentation, build professional PDFs, or needs help authoring markdown intended for the GerdsenAI Document Builder. Handles the full workflow from requirements gathering through markdown authoring to PDF generation. <example>Create a quarterly business review report with charts and tables</example> <example>Write a technical design document for the new authentication system</example> <example>Build a PDF from my project-overview.md file</example> <example>Generate a formatted status report for the team</example>

Adversarial analysis engine that applies multi-domain red team review to code, documents, repositories, security, dependencies, architecture, and infrastructure. Uses Socratic reasoning chains to trace implications across the codebase. Covers code quality, testing gaps, architecture debt, security vulnerabilities, dependency risks, CI/CD issues, database problems, AI/ML concerns, accessibility, and strategic alignment. <example>Red team review this codebase for security vulnerabilities</example> <example>Run adversarial analysis on our API before launch</example> <example>Check this document for unsupported claims</example> <example>Analyze dependencies for CVEs and license issues</example> <example>Full adversarial review of this repo</example> <example>Debug why authentication keeps failing</example>

Use this agent when the user asks to research a topic, create an intelligence report, build a dossier, write a white paper based on research, conduct competitive analysis, perform OSINT research, generate a research-backed report with citations, design a software architecture, plan a tech stack, or research how to build an application. <example>Research the AI chip market and create a report with charts</example> <example>Build an intelligence dossier on quantum computing startups</example> <example>Create a competitive analysis of cloud providers</example> <example>Write a white paper on edge computing trends with citations</example> <example>Design the architecture for a real-time chat application</example> <example>Research how to build a SaaS billing platform</example>

Use this agent when the user wants to plan and execute a development sprint autonomously. Handles Socratic sprint planning, todo.md management, autonomous coding with context-resilient state management, intelligent commit planning, and permission setup. Can write thousands of lines without interruption. <example>Plan and execute phase 1 of the authentication system</example> <example>Complete everything in todo.md</example> <example>Sprint: add payment processing with Stripe integration</example> <example>Resume the sprint from where we left off</example>

Use when authoring, customizing, or extending an AgenticUI agent-observability web app — a generated standalone frontend that shows an AI agent working via a live in-browser terminal (xterm.js + node-pty over WebSocket), a pseudo-desktop (Dockerized Xvfb + x11vnc + noVNC), and optional computer-use (a VLM-driven click/type loop). Activates for the /gerdsenai:agentic-ui scaffolder and any work on its templates.

This skill should be used when the user asks to create a PDF, write a report, build a professional document, generate formatted documentation, author a technical spec for PDF output, or mentions GerdsenAI Document Builder. Covers trigger phrases like 'make me a PDF', 'write a report', 'create a document', 'format this for PDF', 'build a PDF from markdown', and 'generate a formatted report'. Provides rules for YAML front matter, heading hierarchy, code blocks, Mermaid diagrams, tables, and page layout.

Comprehensive meta-skill for discovering and orchestrating ALL available tools, MCP servers, skills, plugins, agents, and CLI capabilities to solve any problem optimally. Invoke this skill for ANY non-trivial task — building software, researching topics, analyzing data, debugging systems, deploying code, creating reports, automating workflows, or solving complex multi-step problems. This skill teaches runtime capability discovery via ToolSearch, parallel agent dispatch, and multi-tool orchestration across Firecrawl (web research), Playwright (browser automation), Pinecone (vector storage/RAG), Vercel (deployment), Cloudflare (edge infrastructure), Mermaid Chart (diagrams), Context7 (library docs), Greptile (codebase search), HuggingFace (ML/AI), CodeRabbit (code review), GerdsenAI Document Builder (PDF reports), language servers (TypeScript, Python, C#, C++), and the superpowers workflow skills (TDD, debugging, planning, code review, worktrees). Even for seemingly simple tasks, run capability discovery first — a specialized tool often produces dramatically better results than a manual approach. When in doubt, use this skill.

3 hooks across 3 events