What problem does it solve? Users often have only a vague, messy, or under-specified idea when they need a complete prompt for a coding agent, research agent, or swarm. This Skill infers the strongest task inside the seed and produces a complete, execution-ready prompt with labeled assumptions, constraints, deliverables, and checkable completion criteria. ## Core Features & Use Cases - Seed-to-prompt forging: Infers the real goal from a rough seed and fills a structured template covering role, goal, assumptions, context, constraints, deliverables, evidence discipline, and done-when criteria. - Dual templates with quality gate: Provides separate templates for repo/code agents and research/writing agents, plus a quality gate checklist every forged prompt must pass before being returned. - Workspace hygiene enforcement: Mandates a verbatim hygiene block for any prompt touching a repository, including read-only-first checks, forbidden-by-default destructive operations, and dirty-worktree quarantine mode. - Use Case: A user pastes "fix the flaky tests in my repo" and receives a full Claude Code prompt with repo context sections, verification commands, a hygiene block, and concrete done-when criteria. ## Quick Start Ask the assistant to forge your rough idea into a master prompt by pasting your seed and saying "forge this into a master prompt for Claude Code".
