What problem does it solve? Rough, vague, or frustrated implementation requests often cause coding agents to produce lazy or incorrect work. This Skill turns under-specified requests into well-grounded, copy-paste-ready prompts by verifying facts first and structuring only what matters. ## Core Features & Use Cases - Pre-investigation Pass: Verifies static facts before writing — files, symbols, GitHub Issues/PRs via gh, configs, and dates — so the refined prompt stands on confirmed evidence, not guesses. - Structured Prompt Output: Produces prompts with Background, Facts, Task, Ideal outcome, and Done condition, plus optional Constraints, Assumptions, and Open Questions. - Lazy-work Guardrails: Strengthens done conditions when a request is likely to produce shallow answers, and requires comparison when multiple solution paths exist. - Use Case: You type "the login flow is broken, fix it" — the Skill checks the actual auth code and related issues, then outputs a precise prompt with file:line references and a clear completion criterion for the next agent. ## Quick Start Ask the agent to refine your rough coding request into a prompt for another coding agent, pasting in your original request as-is.
