LLM Prompt Quality Audit

★ 7.6 · vibe-coding

review-prompt is a Claude Code skill that audits LLM prompts against prompt-engineering gap-calibration principles and reports findings without modifying any files. It first invokes the prompt-engineering skill to load evaluation principles, then produces a structured table of issues with severity levels — High (contradictions, missing goals), Medium (vague directives, restated defaults, boundary failures), and Low (duplication without functional impact) — each tagged AUTO_FIXABLE or NEEDS_USER_INPUT. Only high-confidence issues are flagged; style preferences and unverified hunches are skipped. The skill is designed to run before auto-optimize-prompt, which parses the tags as control-flow signals to decide what it can fix automatically.