What problem does it solve? Ambiguous modification requests often lead to wrong edits, scope creep, or irreversible damage. This Skill establishes change boundaries before any modification happens, separating confirmed facts from assumptions so agents never guess their way through risky edits. ## Core Features & Use Cases - Evidence-First Disambiguation: Inspects code, configs, tests, rules, and history to resolve ambiguity from repository evidence before asking the user anything. - Canonical Intent Resolution: Defines the authoritative policy for when an agent may ask questions—autonomous resolution order, reversible assumptions, single-round questioning, and continuation semantics. - Evidence Gates & Decision Records: Classifies requests as audit, change, or high-risk, and records facts, assumptions, risks, minimal path, and non-goals before medium or large modifications. - Use Case: Before restructuring a repository or editing a shared AGENTS.md rule, run this Skill to confirm the target, affected files, rollback path, and acceptance criteria, asking at most one boundary-changing question. ## Quick Start Use clarify-before-change to analyze the scope, risks, and minimal safe path before refactoring this module.
