Tight, single-task prompts — every change is reviewable and verifiable before the next starts.
You anchor tasks to real files and explicit constraints — Claude doesn't have to guess your codebase.
You delegate the typing, not the thinking — asking for reasoning, not just results.
Average score per dimension across all scored sessions (0–100).
Room to grow in this dimension — coaching details are shown to the profile owner.
Room to grow in this dimension — coaching details are shown to the profile owner.
Room to grow in this dimension — coaching details are shown to the profile owner.
Sign up to see your own full diagnostic with specific coaching on each growth area.
Weakest habit: Verification (25/100) · Plan & Verify
Unverified AI output is the #1 skill-eroding behavior. End every mutating turn with evidence: a test, a build, a curl - something that would FAIL if the change were wrong. The highest-weighted dimension.
→ Holding: Verification is 25 across your last turns vs 26 before. This is the one to push on.
The daily loop: check your one thing → do one rep → come back tomorrow. You're on a 8-day active streak.