Find what's missing from your prompt before Claude guesses it
Every prompt leaves gaps: what's so obvious to you that you don't write it down. Claude fills each gap with a guess, and gets it wrong. Three moves to close them: a blind spot pass, a debrief that pulls the unsaid out of you one question at a time, and a context file Claude re-reads every session. The model is no longer the bottleneck, you are: this kit fixes that.
Why
The model is no longer the bottleneck, you are. You send a prompt that's clear, in your head. Claude hands back something off, and you conclude it didn't understand. In reality, it filled the gaps you left: the target audience, the constraint you find obvious, the format you never named. Every gap is a guess, and one guess in two goes sideways. This kit flips the burden. Instead of hoping your prompt is complete, you surface what's missing (blind spot pass), you get the unsaid pulled out of you one question at a time (debrief), and you fix your context once and for all in a CLAUDE.md. You…
When
For anyone prompting a task they don't master well, or seeing off-target results come back without understanding why. ✅ You know it's your kit when Claude regularly hands back off-target work when your request felt clear to you You tackle an unfamiliar topic and don't even know what questions to ask You repeat the same context (who you are, your stack, your constraints) every new session ❌ Not…
Tools included
- Claude Code
- Prompt Engineering
- Trois Fichiers Projet Ia