What problem does it solve? When someone rearranges a workflow on a visual canvas, the raw edits are ambiguous: a moved card may be layout noise, a reordering, or a tier change, and displaced neighbors are consequences rather than intentions. This Skill reads the change as a structural diff between two graph states and proposes ranked interpretations of intent, so a person can accept, correct, or discard a reading instead of reconstructing it by hand. ## Core Features & Use Cases - Structural diff classification: Distinguishes additions, removals, tier reclassification, reordering, connection changes, reroutes, branches, and orphaned nodes, while marking pure layout moves as non-semantic. - Gesture attribution: Attributes index shifts caused by a single move to the gesture that produced them, so displacement is never reported as separate intent. - Ranked readings with evidence: Ranks intent hypotheses by how well the canvas structure determines them, names the specific structural facts behind each reading, and lists open questions the canvas cannot answer. - Use Case: A user drags a step between two connected steps and removes another. The Skill reports the insertion as a well-determined reading, flags the removal as ambiguous (deletion vs. absorption), and hands off a packet with status needs_clarification rather than writing anything. ## Quick Start Ask the assistant to interpret the recent changes made to the workflow canvas and explain what they imply for the underlying workflow map.
