What problem does it solve? AI agents often dive into non-trivial tasks without externalizing the goal, success criterion, or likely failure mode, leading to wasted tokens, silent wrong assumptions, and work committed to the wrong branch. This Skill enforces a short, evidence-backed thinking opener before any tool calls. ## Core Features & Use Cases - Six-line structured opener: goal with decision criterion, ordered approach bullets, scope with unverified assumptions, one-line premortem, named verifier, and branch/worktree routing decision. - Auto-trigger rules: fires on tasks with 2+ tool calls, decisions under uncertainty, research or audit work, and cross-file edits, while explicitly skipping mechanical one-line fixes and QA runs. - Inferential check for strategic tasks: tags load-bearing claims by evidence level, names biases to guard against, and requires a one-line falsifier on quantitative or causal reasoning. - Use Case: Before refactoring across multiple files, the agent writes the opener first, catching an unstated assumption about scope with a quick grep and confirming the correct target branch before any code touches disk. ## Quick Start Ask the agent to plan a non-trivial task and it will write the six-line think-better opener covering goal, approach, scope, premortem, verifier, and routing before making any tool calls.
