An Agent Skill that makes things click: it turns AI output into the format a human understands fastest inspired from Andrej Karpathy's X Post
Python
0
1 commits
updated Oct 5, 2026
See what people are saying
README
make-it-click
An Agent Skill that makes things click: it turns AI output into the format a human understands fastest inspired from Andrej Karpathy's X Post.
As models do more of the work, people spend more time checking and understanding that work. This skill treats the format of an answer as a choice. It picks the lowest step on a ladder that does the job, and moves up when the content needs it:
- Controlled writing in the style of ASD-STE100 (Simplified Technical English): short sentences, active voice, one meaning per word. Strict, "80%", or light.
- Diagrams: flowcharts, architecture, timelines, one-page reference sheets.
- Interactive HTML pages: self-contained, step-through, sliders, before/after.
- Narrated explainer videos: Manim animation + ElevenLabs or free local TTS.
It also has an oversight mode for reviewing agent work: what changed → why → risk → what the human must check.
Install
Claude.ai / Claude Desktop: download make-it-click.skill from Releases and add it
under Settings → Capabilities → Skills (or upload the skill/make-it-click folder as a zip).
Claude Code: copy skill/make-it-click into ~/.claude/skills/ (personal) or
.claude/skills/ in your project.
Try it
- "Explain how TCP congestion control works, 80% STE."
- "Diagram what this agent changed in the repo and what I need to check."
- "Make an interactive HTML explainer of gradient descent."
- "Create a 3b1b-style video explainer on Fourier series. Use local TTS."
STE checker
python skill/make-it-click/scripts/ste_check.py my_text.md
A heuristic linter. It flags long sentences, -ing forms, passive voice, perfect tenses, and common unapproved words. It is not a compliance tool.
Credits and notices
- The idea of the format ladder comes from public notes by Andrej Karpathy on making LLM outputs easier to understand. This project is not affiliated with or endorsed by him.
- ASD-STE100 is a specification maintained by ASD (Aerospace, Security and Defence Industries Association of Europe). This repository does not contain the specification or its dictionary. The rule summary here is written independently and is not authoritative. Get the official specification from https://www.asd-ste100.org. "ASD-STE100" is used only to refer to that specification.
License
MIT. See LICENSE.
Similar repos
An Agent Skill that makes things click: it turns AI output into the format a human understands fastest inspired from Andrej Karpathy's X Post
Python
0
1 commits
updated Oct 5, 2026
See what people are saying
README
make-it-click
An Agent Skill that makes things click: it turns AI output into the format a human understands fastest inspired from Andrej Karpathy's X Post.
As models do more of the work, people spend more time checking and understanding that work. This skill treats the format of an answer as a choice. It picks the lowest step on a ladder that does the job, and moves up when the content needs it:
- Controlled writing in the style of ASD-STE100 (Simplified Technical English): short sentences, active voice, one meaning per word. Strict, "80%", or light.
- Diagrams: flowcharts, architecture, timelines, one-page reference sheets.
- Interactive HTML pages: self-contained, step-through, sliders, before/after.
- Narrated explainer videos: Manim animation + ElevenLabs or free local TTS.
It also has an oversight mode for reviewing agent work: what changed → why → risk → what the human must check.
Install
Claude.ai / Claude Desktop: download make-it-click.skill from Releases and add it
under Settings → Capabilities → Skills (or upload the skill/make-it-click folder as a zip).
Claude Code: copy skill/make-it-click into ~/.claude/skills/ (personal) or
.claude/skills/ in your project.
Try it
- "Explain how TCP congestion control works, 80% STE."
- "Diagram what this agent changed in the repo and what I need to check."
- "Make an interactive HTML explainer of gradient descent."
- "Create a 3b1b-style video explainer on Fourier series. Use local TTS."
STE checker
python skill/make-it-click/scripts/ste_check.py my_text.md
A heuristic linter. It flags long sentences, -ing forms, passive voice, perfect tenses, and common unapproved words. It is not a compliance tool.
Credits and notices
- The idea of the format ladder comes from public notes by Andrej Karpathy on making LLM outputs easier to understand. This project is not affiliated with or endorsed by him.
- ASD-STE100 is a specification maintained by ASD (Aerospace, Security and Defence Industries Association of Europe). This repository does not contain the specification or its dictionary. The rule summary here is written independently and is not authoritative. Get the official specification from https://www.asd-ste100.org. "ASD-STE100" is used only to refer to that specification.
License
MIT. See LICENSE.