What problem does it solve?
Exploring an unfamiliar codebase is slow and error-prone when you rely only on manual file reading. This Skill guides an AI agent to use a Codebase Memory MCP server's indexed code graph for fast discovery of symbols, architecture, and dependencies, while enforcing evidence standards so graph results are verified against actual source before making claims.
Core Features & Use Cases
- Graph-Guided Discovery: Use
search_graph, get_architecture, and trace_path to find definitions, implementations, callers, callees, data flow, and dependency paths across a repository.
- Tiered Evidence Levels: Apply Scout, Verify, or Auditor evidence standards so negative, exhaustive, dead-code, and impact claims are backed by index freshness checks, coverage inspection, and source confirmation.
- Safe Indexing and Fallbacks: Control indexing modes (fast, moderate, full), require explicit approval before mutations like
index_repository or delete_project, and fall back to rooted filesystem exploration when the graph is unavailable.
- Use Case: When asked whether removing a function is safe, the agent traces all callers through the graph, checks index coverage for every cited path, paginates complete result streams, and confirms with source snippets before concluding the function is dead code.
Quick Start
Ask the agent to map the architecture of this repository and trace all callers of a specific function using the Codebase Memory graph, verifying each finding against the actual source files.
