Indexes Git repositories to semantically navigate code and find cross-file references for Ruby, Java, Kotlin.
npx skills add https://github.com/hidonguyen/trade-ops --skill ck-gkg-hidonguyen
Please help me install this Agent Skill. Skill: ck:gkg Source: https://github.com/hidonguyen/trade-ops/tree/main/.opencode/skills/gkg Command: npx skills add https://github.com/hidonguyen/trade-ops --skill ck-gkg-hidonguyen
Large repositories and polyglot codebases make it hard to find definitions, usages, and the impact of changes across files and projects; traditional text search and ad-hoc grep are noisy and error-prone. This Skill provides a queryable semantic index that enables precise go-to-definition, find-usages, and impact analysis to reduce refactoring risk and speed code comprehension.
Index the current repository and start the local GKG server so you can query definitions and find usages through the HTTP or MCP endpoints.
High-intent search queries and answers about installing and using this skill.
To perform semantic code navigation across large repositories, you must build a persistent AST-based knowledge graph. This index enables precise go-to-definition and find-all-usages queries to reduce refactoring risk and speed code comprehension.
Impact analysis across multiple files is best done by querying a semantic knowledge graph to identify call sites and dependency graphs. This approach replaces noisy text search and ad-hoc grep with precise cross-file reference discovery before refactoring.
Semantic code navigation works with Ruby, Java, and Kotlin repositories that have strong AST support. It also provides partial support for Python, TypeScript, and JavaScript to help map polyglot codebases.
Querying cross-file references requires an initialized Git repository, a built gkg index, and a running local server. The server exposes REST and MCP endpoints using persistent KuzuDB storage under your ~/.gkg workspace.
AI assistants can use local HTTP APIs and MCP tools to list projects, read definitions, and get references. These integration points allow assistants to produce token-efficient repository maps for architecture reviews.
Text search fails in large codebases because it is noisy and error-prone when matching identical strings across different contexts. A queryable semantic index provides precise cross-file reference discovery instead of fragile pattern matching.
phuongrealmaxEnable semantic code search across large repositories with tree-sitter chunking and knowledge graphs.
phanijappsMap software architecture and trace code graphs using the engram MCP.
Endor-Solutions-ArchitectureQuery Endor Labs findings, projects, packages, and OSS metrics via API endpoints.
massgenLocate text and structural code patterns across large codebases using ripgrep and ast-grep.
