What problem does it solve? AI agents working on the VitTrade Flutter repo waste tokens and get wrong answers when they misuse the memory-for-ai MCP knowledge graph, such as using it for exact-name searches or trusting it as the source of truth for edits. This Skill encodes measured decision rules so the graph is used only where it beats grep and Read. ## Core Features & Use Cases - Blast-radius analysis: Trace inbound callers and detect change impact before editing shared widgets, spacing tokens, scaffolds, or the router, where grep cannot answer multi-hop impact in one call. - Architecture and discovery queries: Use get_architecture, query_graph with complexity columns, and BM25 or semantic search when no exact symbol name is known. - Accuracy guardrails: Enforce freshness checks against git log, textual-resolution caveats for Dart edges, and safe retry patterns for connection errors. - Use Case: Before refactoring a Vit shared widget used by dozens of features, run trace_path inbound to list impacted callers and include them in the edit plan. ## Quick Start Before editing any shared widget or router file, ask the agent to check the memory-for-ai graph for callers and blast radius following this Skill's rules.
