forge-promptlisted

# forge-prompt Turn a rough request into a precise, buildable instruction. The goal is to improve the *quality of the thinking* behind the prompt — not to wordsmith. Output a sharpened prompt the user approves, never a silent reinterpretation. ## Argument Text after the command is the rough request to sharpen (e.g. `/forge-prompt make the dashboard faster`). No argument → ask for the request to optimize. ## Principle The user owns intent. This skill makes intent explicit and complete; it does not invent goals, change scope, or assume requirements. When something is ambiguous, ask — don't guess. ## Steps ### 1. Read the raw request and the project context Take the user's rough request. If the project uses the methodology (a `context/` folder exists), read `project-overview.md` and `architecture.md` for scope and invariants, and the build plan / specs if relevant. This grounds the optimized prompt in the real system instead of generic advice. ### 2. Detect gaps Identify what's missing for a high-quality prompt: - **Goal** — is the concrete outcome clear? - **Scope** — which unit / boundary / files does this touch? Anything explicitly out of scope? - **Constraints** — invariants, tech choices, patterns from the context files that apply. - **Acceptance** — how will "done" be verified? ### 3. Ask only what's necessary If critical gaps exist, ask 1–2 focused clarifying questions. Do not interrogate — ask the fewest questions that remove real ambiguity. If the request i