What problem does it solve? Vague or rough task requests cause AI agents to invent assumptions, skip verification, and produce unverified or unsafe work. This Skill converts an ambiguous request into a structured GOAL template with explicit facts, constraints, verification steps, and stop rules before any execution begins. ## Core Features & Use Cases - Pre-GOAL Fact Pass: Gathers cheap, fetchable facts from code, logs, issues, and docs before asking the user anything, separating confirmed facts, blocked facts, default decisions, and genuine uncertainties. - Decision Funnel: Runs a question-driven decision ledger (confirmed, likely, autonomous, open, out-of-scope) for product, design, or spec-shaping work before writing the GOAL. - Risk-Calibrated Templates: Classifies tasks by type and risk level, producing compact templates for small tasks and full templates with STOP RULES and HITL approval loops for high-risk work like security, production, or external posting. - Use Case: You have a vague request like "investigate whether WAF is blocking staging traffic and fix it." The Skill inspects available evidence, asks only the decisions that matter, and outputs a copy-paste-ready GOAL prompt another agent can execute without re-litigating settled decisions. ## Quick Start Ask the agent to turn your rough task description into a GOAL template for handoff to another agent.
