Introducing Skills on ConsoleX AI

Today we’re shipping one of the biggest upgrades to ConsoleX AI since our “agentic tool use” launch: Agent Skills.

Skills: A Better Foundation for Agentic Work

ConsoleX AI previously offered an autonomy-mode workflow: you enabled tool calling, and the model determined when to act. While powerful, long-running agentic tasks frequently encounter friction—inconsistent planning, instruction drift, tool misuse, or context bloat.

Agent Skills—based on the standard introduced by Anthropic that has rapidly become the norm for AI agents—fundamentally reinforce ConsoleX AI, equipping the platform with the structural backbone needed for reliable agentic systems. This upgrade transforms scattered prompts and fragile autonomy into reusable, dependable building blocks. By attaching a Skill, ConsoleX AI allows LLM to operate with greater stability, higher consistency, and reduced token overhead. Since Skills now encapsulate Tool/MCP associations, the platform enables autonomous orchestration of the right tools without repetitive manual setup.

In other words: Skills turn “do it autonomously” into “do it autonomously, but with a proven playbook.”

A Cloud AI Studio Built for Skills

We believe ConsoleX AI may be among the first cloud-based AI studios to support Agent Skills end-to-end—designed not only for using prebuilt skills, but also for creating and uploading your own skill packs, and binding skills to Tools/MCP servers.

Prebuilt Skills: Automation Without Tool Plumbing

Prebuilt skills seamlessly bridge ConsoleX AI's capabilities, orchestrating tools organically and loading them progressively on demand. This approach eliminates setup friction while covering common agentic workflows, including:

• artifact-builder — Build self-contained visual artifacts (React components, HTML pages, SVG, Mermaid diagrams) that benefit from rendering/preview.
• cloud-browser — Operate a cloud Chrome browser for web automation (navigate, click, fill forms, screenshots, multi-step flows).
• code-execution — Run Python in a secure sandbox (calculations, data analysis, charts, testing code).
• doc-to-slides-images — Turn a document/URL/text into a NotebookLM-style slide deck as PNG images (typically 8–12 slides, up to 20).
• document-suite — Read/write Office docs and PDFs: DOCX, XLSX, PPTX, PDF (extract, create, edit, merge/split, etc.).
• domain-name-brainstormer — Generate domain ideas and check availability across TLDs (.com, .io, .dev, .ai, etc.).
• frontend-design — Produce polished, production-grade UI/front-end code and designs (less “generic AI” look).
• multimedia-ai-generation — Generate/edit images, generate videos, and generate audio (supports multiple models; tool-based generation).
• url-file-parser — Fetch and parse content from public URLs (web pages or downloadable files; routes to the right parser).
• web-search — Real-time web search via Brave for up-to-date info, verification, current events.
• youtube-transcript — Extract full transcripts/captions from YouTube videos (with or without built-in subtitles).

Custom Skills: Native Support for Open Agentic Skills

ConsoleX AI's capability is significantly expanded by its native compatibility with the Agent Skills standard. By adopting the standard directory structure (SKILL.md, reference/, scripts/), we allow you to deploy any compliant skill package directly to the cloud. This transforms ConsoleX AI into an extensible platform where you can instantly plug in open-source capabilities or specialized domain logic, without the friction of custom integration or manual setup.

If you want a guided path, the skill-creator skill helps you draft a new skill from scratch: define triggers, write the workflow, and structure reusable resources so the skill stays lean and loads progressively.

Safe Script Execution with E2B Sandboxes

Script execution is a vital component of Skills. We separate this process into two distinct layers:

• Decision Making: The model decides whether to act and what action to take (outputting structured JSON for a Tool Call).
• Execution: The system executes the action in a controlled environment and feeds the results back to the model.

For security, all script execution in ConsoleX AI takes place within an isolated environment. We leverage E2B sandbox to dynamically provision sandboxes for script execution. A dedicated tool, sandbox_exec, invokes the sandbox to handle operations—such as file I/O, script running, format conversion, and artifact storage—creating a reliable "security fence" around code execution.

Skills + Tools/MCP: Make the Model the Orchestrator

Skills become much more powerful when combined with ConsoleX’s Tools and MCP ecosystem. You can bind a skill to specific tools and MCP servers, run deterministic steps in the sandbox, and keep a human in the loop for style/content decisions—while the model “conducts” the multi-step tool calling.

This release also includes:

• Parallel conversations (start a new chat while another one is still running)
• A redesigned interface that is cleaner and more intuiative to use
• A new My Stuff side dock (Files, Prompts, Skills) with drag-and-drop assets into chat
• Starred conversations for quick access
• A faster, more flexible Assets workflow for generation and remix

Parallel Conversations: No More Waiting

Long-running agentic tasks are great—until they block your workflow.

Previously, you could have only one active “running” conversation at a time. Now, ConsoleX AI supports parallel conversations: start a new chat immediately, even if another conversation is still running.

A New, Cleaner Interface

We’ve redesigned the main chat experience to be simpler and clearer. A unified “Add” button now integrates file uploads, tool selection, and deep-thinking modes into a single, logical entry point. This streamlined design reduces UI friction and makes the interface significantly easier to use, allowing you to focus entirely on building and running agentic workflows.

My Stuff side dock

A new My Stuff side dock brings your Files, Prompts, and Skills into one place.

Starred Conversations: Pin What Matters

You can now star any conversation and treat it like a living workspace.

Faster Asset Generation and Remix

The Assets panel is now a faster creative loop:

• Click Generate to produce images, audio, or video content, and freely switch models without restarting your flow.
• Image previews are now an interactive canvas: migrate styles, remix variations, edit specific regions, or even generate a video from an existing image.

This makes it easy to go from “idea” → “first draft” → “remix” → “final” without leaving ConsoleX.

Want to try Skills? Open a chat on ConsoleX AI, enable Skills, and let your agent run with a playbook.

— The ConsoleX Team