csp-code-graphlisted

# Code Knowledge Graph — Methodology & Lifecycle A structural knowledge graph of the codebase enables **token-efficient, context-aware** code reviews. Instead of re-reading entire files, the graph tells you exactly which functions, callers, dependents, and tests are affected by a change — achieving 38x–528x token reduction over naive file scanning. ## Core Principle > **Read what matters, skip what doesn't.** The graph maps structural relationships (calls, imports, inheritance, tests) so review effort concentrates on the blast radius of changes rather than scanning the entire codebase. ## Graph Model **Node types:** File, Class, Function, Type, Test **Edge types (with traversal weights):** | Edge | Weight | Meaning | |------|--------|---------| | CALLS | 1.0 | Function invokes another function | | INHERITS | 0.9 | Class extends another | | IMPLEMENTS | 0.9 | Class implements interface | | INHERITS (cross-lang) | 0.8 | Cross-language inheritance | | TESTED_BY | 0.7 | Function covered by test | | DEPENDS_ON | 0.6 | General dependency | | REFERENCES | 0.6 | Value-level reference (callbacks) | | IMPORTS_FROM | 0.5 | File imports from module | | TESTED_BY (indirect) | 0.4 | Indirect test coverage | | CONTAINS | 0.3 | Structural containment | **Confidence tiers:** EXTRACTED (from AST) > INFERRED (from heuristics) > AMBIGUOUS (uncertain) ## Lifecycle Stages The code graph operates through a lifecycle. Route to the appropriate skill: ``` ┌──────────────────────────────────