claude-code-iterate

Prompt enhancement skill. Takes a vague task description, identifies what's missing or ambiguous, rewrites it into a structured prompt using prompt engineering best practices (specificity, chain-of-thought, role context, output format), generates testable success criteria, and executes with built-in verification. Turns "build me a dashboard" into a complete spec.

Self-correcting agentic loop. Executes a task, evaluates output against success criteria, and retries with targeted fixes until all criteria pass or max iterations reached. Use when a task needs to meet specific quality bars, pass tests, match a spec, or satisfy multiple requirements that are unlikely to all be met on the first attempt.