What problem does it solve? Working on lci's function purity and side-effect analysis requires navigating a large C++ codebase spanning AST extraction, call-graph propagation, MCP handlers, and annotation parsing, with many invariants and known traps that are easy to violate. ## Core Features & Use Cases - Code Map: Pinpoints every surface and file:line entry for the side_effects and semantic_annotations MCP tools, the purity block in get_context, and the analyzer pipeline from tree-sitter extraction to transitive fixpoint propagation. - Invariants and Traps: Documents conservative-classification rules, determinism requirements, sink wiring order, known gaps (Go closures, stale incremental records, dead category flags), and doc drift issues. - Probe Recipes and Tests: Provides ready JSON-RPC probe commands, targeted gtest filters, and a before/after corpus-diff recipe for classifier changes. - Use Case: When changing the callee-category keyword tables, use this Skill to find classify_callee_category, run the word-boundary tests, and diff classification output against a real corpus in both directions. ## Quick Start Ask the AI to locate the purity and side_effects code paths in the lci repo and explain how transitive impurity propagation works before modifying the classifier.
