AI column

Docs & writing@donghyeok99Staff

Answer me with HTML: agent answers as one HTML page

In the README's runs, a model asked to hand-write the HTML took 31 s; through this skill the same answer page took 12 s

Instead of 4,893 tokens of HTML, the model writes a 612-token Markdown draft; about 50 ms later the CLI draws the TCP handshake on a page.
Instead of 4,893 tokens of HTML, the model writes a 612-token Markdown draft; about 50 ms later the CLI draws the TCP handshake on a page.
Who
An MIT-licensed skill by QingYunA, for anyone who finds an agent's long answers hard to take in.
When
Published on October 2, 2026; the latest release is v0.5.10 from October 26.
Where
Claude Code, Codex, Cursor, OpenCode, Pi and DeepSeek Harness. Needs Node.js 20+, runs on your machine, pages open offline.

What was hard

Ask an agent how something is structured and a wall of text scrolls past in the terminal, and you rebuild the picture in your head. Ask it for HTML and the page is decent, but each one took a minute or two, mostly hundreds of lines of CSS that were nearly the same every time. Diagrams were worse: the model worked out SVG coordinates by hand, and arrows often pointed at nothing.

What it does about it

The model writes only the content, as a short Markdown draft, and the am CLI bundled with the skill handles layout, colour and drawing. You list relations in a sequence or flow block; the CLI computes the coordinates, applies a theme and writes one HTML file with no dependencies. Across nine pages the README measured, hand-written HTML averaged 4,893 tokens and the draft for the same question 612, about one eighth.

The 24-second demo: from the question to the open page
One TCP question: terminal text on the left, the skill's page on the right
One TCP question: terminal text on the left, the skill's page on the right
The TCP handshake page made from one draft, examples/tcp.en.md
The TCP handshake page made from one draft, examples/tcp.en.md
Four frames of the explainer video you get when you ask for a 3b1b-style one
Four frames of the explainer video you get when you ask for a 3b1b-style one
> Explain the TCP three-way handshake> Map out how the modules in this repo fit together> Redis or Memcached for our cache?
The README's own run

What people made with it

GitHub@tpanagos2026-10-19

Tim Panagos of Vibe Rooster ran am on the plain weekly ops page his collection uses as its 'before', and published the after page: panels of key numbers, next actions and a flow chart.

I ran `am` on the plain weekly ops page we use as the shared "before" for every skill in the collection, and published the result
likes 0 · comments 1 (2026-10-29 as of)GitHub ↗

How it runs inside

  1. 1

    You → Claude Code, Codex and others

    Explain the TCP three-way handshakeYou ask as you always do, with no special command.
  2. 2

    SKILL.md → Claude Code, Codex and others

    Whether to answer with a page, and which componentsThree or more related concepts, a flow or a comparison means a page; anything else gets plain text.skills/answer-me-with-html/SKILL.md
  3. 3

    Claude Code, Codex and others → am CLI (am.mjs)

    am render - <<'AM_EOF' (the Markdown draft)The model writes no HTML, CSS or SVG, only panels with sequence and flow blocks.src/cli.js
  4. 4

    am CLI (am.mjs) → Claude Code, Codex and others

    ✗ L<line> + Correct example, STE warningsThe draft is parsed and checked against rules adapted from ASD-STE100; errors come back with a line and a fix.src/lint/ste.js
  5. 5

    am CLI (am.mjs) → dagre

    The A -> B relations of a flow blockdagre computes the coordinates from the relations alone, and flow.js draws that layout as SVG.src/components/flow.js
  6. 6

    am CLI (am.mjs) → One HTML file in pages/

    .html with theme, reply script and source draftThe panels go into a template, CSS and runtime are inlined, and the draft is kept inside one file.src/render.js

What goes in, what comes out

Explain the TCP three-way handshakeA sequence diagram, a state diagram and a flag table
How are the modules in this repo organized?A folder tree plus a call graph
Redis or Memcached?A comparison table with ✓ and ✗, then a verdict
How do I show hidden files with ls?No page, just a one-line answer

Is it for you

Good for

  • You'd rather read an agent's long explanations as diagrams and tables
  • You often ask about repo structure, protocol flows or trade-offs between tools
  • You want to keep or pass on an answer as one file that opens offline
  • You want to pick options and comment on a page, then send it all back to the agent as one reply
  • You want to try 3Blue1Brown-style explainer videos

Not for

  • Needs Node.js 20 or newer.
  • The bill drops less than the writing: every turn still reads the system prompt and the conversation, about 15% less in the README's runs.
  • Exporting a video as MP4 needs ffmpeg; WebM works without it.
  • Updates are manual, and pages pile up in ~/.answer-me-with-html/ until you run am clean.

Reviewed by Devalley’s editor before publishing. How the column is made

Sign in to post, comment and like.Sign in

Report

Reason

Read the community rules