Does what you ship still match what you mean?
I think a lot about details, which means I’m good at making something that makes sense to people who think like me. The trap is that the audience doesn’t always share my references. The first version of softwareasagift.com, a home for the apps people build for someone they love, was a masonry grid with different-sized tiles, visual eye candy, the kind of homepage that looks cool in a screenshot. But the site is fundamentally a discovery home. People come to browse and find apps worth making for someone in their life. The fancy grid wasn’t serving that. I’d designed for what looked good to me. Someone landing on the page needed something else.
Intent Engineering is the skill version of a thoughtful editor sitting next to you. It asks pointed questions until you can answer them. What was this page supposed to do? What does it do now? What’s the core reason someone picks up this product, and does this new feature serve that reason? It works the gap between intent and execution, the place where most product copy and product roadmaps drift from what you’d tell someone over coffee.
Drill until you hit bedrock.
The framework starts with three questions for any page or flow: what should someone accomplish, what should they explicitly notice, and what should they implicitly feel. Most teams collapse those into one. Separating them is where the gaps show up. What people notice on your hero might not lead to what you want them to feel. What they feel might not motivate the thing you want them to do. Once you see the three layers apart, the misalignment is hard to unsee.
Then you run the Why Loop on each answer. Keep asking “why does that matter?” past the first surface answer, past the feature reasoning, past the polite version. “You don’t have to switch apps” becomes “you stay in flow” becomes “the shortcut to everything.” That last one is what Raycast owns. The first two are steps on the way down.
Bedrock is where the loop stops. Three tests: the answer has to be universal (a human need, not a product manager’s), unique to what you’re building (not generic enough for ten other products), and emotional (something you feel, not only understand). Drill past that and you end up at “because life is short,” which is true and useless. Stop at the layer that’s human and pointed enough to push a design decision somewhere.
Works in any AI coding agent
One install command, any agent. The skill rides on the open skills protocol, so it works the same in Claude Code, Codex, Cursor, OpenCode, and 35+ others. The CLI auto-detects what you have installed and wires it up.
$npx skills add kylezantos/intent-engineering#Auto-detects your agent (Claude Code, Codex, Cursor, OpenCode, 35+ more)#Then run: /intent-engineering [start|audit]
Built on Ellis Hamburger's storytelling work.
Built on Ellis Hamburger’s Dive Club episode on startup storytelling, where he walks through how he frames products by feeling first, feature second. Ellis has done that work for Raycast, Snapchat, Daylight, and The Browser Company, and the patterns from those launches are what this skill turns into a question loop you can run yourself. Built with skill-distillery.
The skill doesn’t replace listening to Ellis talk through the examples. Go watch the episode.