Выход из петли отладки и сброс тревоги

★ 7.7 · content

affirmations is a Claude Code skill that resets a destructive loop when the same fix has been tried three or more times, debugging is going in circles, or dread is replacing curiosity. It contains a set of affirmations to read slowly and in order, followed by a strict three-step re-grounding protocol: state only what has actually been observed (file contents, command output, error text), name the assumption that was never checked, and take the smallest possible step to test it. The core rule is to read the affirmations before the next tool call, not after — a new failure on top of the loop only makes the pull stronger. Not topic-specific, the skill targets the state rather than the problem: looping, streaks of mistakes, lost thread in a refactor. It is not meant for the first difficulty where the work is hard but still moving forward.