prompter

What problem does it solve? Short, grammatically fine prompts are often under-specified along the dimensions that matter, so Claude Code runs in a direction you didn't choose and the next turn is spent reorienting. This Skill intercepts prompts before execution, evaluates whether a rewrite would meaningfully improve them, and shows the rewrite for approval first. ## Core Features & Use Cases - Evaluate-first rewriting: Only rewrites when there is a real improvement — typos, ambiguous references like "that file", vague intent, or missing guardrails — and emits a skip notice otherwise. - Approval-gated rewrites: Displays the rewritten prompt with a "Rewritten prompt:" prefix and waits for your approval before acting. - Flexible activation modes: Try it once, apply for the whole session, or persist the behavior by adding a Prompt Rewriting rule to your project's CLAUDE.md (with safe detection, diffing, and removal of existing sections). - Use Case: You type "the tests are flaky, fix them" and the Skill rewrites it into a structured diagnosis task — categorize each flake by type, produce a ranked table, and don't apply fixes yet — before any code is touched. ## Quick Start Ask Claude to run the prompter skill, choose "Try it once", then submit a short prompt like "the tests are flaky, fix them" to see the proposed rewrite.