- Home
- Skills
- Agents & Orchestration
- Explaining Visually
Explaining Visually
Plans and walkthroughs as a local, commentable web page instead of a wall of text: comment on any box, the agent edits and replies live.
Free
Works with the AI Agents you already use
Explaining Visually
Example session with this skill installed
Make a plan for moving session tokens into httpOnly cookies.
- Read your context and instructions
- Compiled the explaining visually
Summary
The agent wrote the plan as a page instead of a chat answer, started the local explainer server and is now listening for your comments.
Page: http://127.0.0.1:4791/2026-10-22-cookie-auth/
What the page shows
Questions for you (band on top, one at a time, with a recommended option and a free-text field)
- Should the mobile app move to cookies too? Recommended: Move it too
- Does any third-party integration read the token from the login response body?
| Section | Content |
|---|---|
| TL;DR | Why: an XSS bug today means a stolen session. What changes: login sets a cookie. Size: 3 files, no schema change. |
| Today | Login form → token in body → localStorage (auth/store.ts:14) → Authorization header → API |
| After | Before/After columns: localStorage write removed, login response changed to Set-Cookie; HttpOnly; SameSite=Lax, cookie middleware new |
| Trap | SameSite=Lax still sends the cookie on top-level GETs, so every GET must stay side-effect free |
| Steps | 1. Cookie middleware beside bearer auth 2. Login sets the cookie 3. Drop localStorage and the bearer path |
| Where | api/auth/cookie.ts, api/routes/login.ts, web/auth/store.ts |
Every file:line opens the real code in a side sheet.
Next steps
- Answer the two questions, or click any box and comment on it.
- Each comment reaches the agent with the exact source line; it edits only that element, replies, and the page outlines what changed.
- When it looks right, press Approve plan and the agent starts step 1.
Connects securely to your tools. The creator never sees your data.
About this skill
Technical explanations often collapse into walls of text that stakeholders ignore or misunderstand. This skill converts architecture walk-throughs, design proposals, and code reviews into interactive, commentable visual explainers served from a local environment.
What it does
- Visual scaffolding generates HTML fragments using structured components like flows, before/after grids, and commit-sized steps.
- Interactive commenting enables users to click any element in their browser to send feedback directly back to the agent via monitor events.
- Live plan negotiation uses structured question blocks (
.q) to resolve open decisions before finalizing implementation details. - Local serving launches a background node server to host explainers for the current repository, accessible via local or remote URLs.
- Contextual expansion hides deep technical details or code examples in a side sheet to keep the primary view skimmable.
How it works
- Proposal triggers a prompt asking if the user wants a visual explainer for the current topic.
- Generation writes a
.body.htmlfragment to a dedicated local directory using specific layout classes like.flowor.steps. - Review starts a background server and a file watcher to capture browser-based comments as monitor events.
- Iteration updates specific line numbers based on feedback and sends replies to individual comment threads.
Frameworks & tools
Uses Node.js for the local server and watcher scripts. Generates GitHub-flavored Markdown for approved plans and standard HTML/SVG for visual components.
Why this beats prompting it yourself
Manual prompting rarely produces consistent visual structures or manages the complex loop of linking browser comments to specific lines of code. This skill automates the feedback cycle, ensuring every suggestion is tracked, replied to, and applied without manual copy-pasting.
Use cases
- Proposing a complex refactor with a before/after visual comparison.
- Walking a new developer through the mental model of a specific subsystem.
- Resolving design decisions in a plan before writing any production code.
- Creating skimmable, interactive documentation for a specific feature's logic.
Known limitations
Does not support cloud-only sessions where a local port cannot be exposed. Requires a local Node.js environment to run the server and watcher scripts.
Requires local Node.js environment. Cannot serve pages in cloud-only environments. Browser interaction requires a local or tunneled connection to the agent machine.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
1 use
Times used to date
Free forever
No account required to browse
Trust & safety
Security scan
Security scan passed 25 days ago
- Free with an Agensi account
Requires