github.com/QingYunA/answer-me-with-html
- 1.5K+ GitHub stars (1,501 on 2026-10-06 UTC), MIT license
- Author's benchmark vs asking the agent to write the HTML by hand (Claude Sonnet 5.5, 3 topics x 3 runs, medians): 5,341 -> 870 output tokens (6.1x fewer), 33 s -> 12 s (2.8x faster), $0.092 -> $0.067 per answer, in a plain Claude Code setup
- In a heavy setup (about 51,000 tokens of tools, rules and skills) the author measured the skill about 20% more expensiveabout 51K tokens
- Needs Node.js 20+CLI bundled inside the skill, nothing else to install
Verified Oct 5, 2026
What it is
A free, MIT-licensed skill for coding agents (Claude Code, Codex, Cursor, OpenCode and more). Ask a hard question and the agent writes a short Markdown draft; a CLI bundled in the skill turns it into a single HTML page: sequence diagrams, flow charts, ✓/✗ tables, timelines. A one-line question still gets a one-line answer, no page.
Setup
You need Node.js 20 or newer. Nothing else to install: the CLI is bundled inside the skill.
Easiest: paste this into your agent (Claude Code, Codex, Cursor, OpenCode…):
Install Answer me with HTML: read https://raw.githubusercontent.com/QingYunA/answer-me-with-html/main/INSTALL.md and follow it.Claude Code plugin instead: run these inside Claude Code.
/plugin marketplace add QingYunA/answer-me-with-html/plugin install answer-me-with-html@answer-me-with-htmlAny other agent, one command (the installer asks which agents to use):
npx skills add QingYunA/answer-me-with-htmlAsk a question the way you always do, for example "Explain the TCP three-way handshake". Pages are saved in
~/.answer-me-with-html/pages/.
Cheat sheet
- Repository and README.
- Agent-friendly install guide.
- Benchmark and the script to reproduce it: the 6.1x and 2.8x are the author's own medians (3 topics x 3 runs, Claude Sonnet 5.5, plain Claude Code setup), measured against asking the model to write the full HTML (not against a plain-text answer), and not an independent test.
- Heavy setups: the skill adds two short turns and every turn re-reads your context. With about 51,000 tokens of tools, rules and skills loaded, the author measured it about 20% more expensive. Speed-up held in every setup they tried.
- Pages open in your browser after they are made. Turn that off with
/answer-me-with-html config open off(plugin install:/answer-me-with-html:config open off). - Always-on mode (off by default, and the README now recommends it): add one rule to your agent's rules file and every conclusion gets a small page. Pages pile up in
~/.answer-me-with-html/; clear them witham clean. To turn it off, delete the rule. - Explainer videos (
am video): narration uses ElevenLabs only if you setELEVENLABS_API_KEY, otherwise the system voice. The MP4 export needs Chrome, ffmpeg and Node.js 22+.
Axo's take
Haven't run it myself: this page comes from the project's README and install guide. Try it if your agent keeps answering with a wall of text. The numbers are the author's, so check them against your own setup.
Get every link first
Follow me for the next drop. Axo's Den, my broadcast channel, opens soon. Free. No spam. I'm an axolotl, I don't have the energy.

