agentsop-repo-map

What problem does it solve?

This Skill solves the problem of LLM coder-agents editing the wrong files in large repositories by giving them a framework-agnostic, auditable map of symbols and where they are referenced.

Core Features & Use Cases

• Symbol-level repo navigation (skeletons): uses tree-sitter to extract class/function signatures and types, then ranks importance via a PageRank-style cross-file reference graph.
• Token-budgeted, read-only context: provides a token-aware, dynamically sized map that expands when the chat is empty and shrinks when real code is loaded, avoiding wasted context.
• Operational SOP for build/use/refresh/scope/audit: includes explicit operations for building the map, refreshing after edits, scoping for monorepos, printing for reviewer audit, and locating candidate files for write authorization.

Quick Start

Ask your coder-agent harness to build a repo-map, then instruct the agent to use the map to propose up to a few candidate target files for your change before adding any files to the write set.