Build and maintain dual-layer dependency and concept graphs for gk-stack projects.
npx skills add https://github.com/gkganesh12/gk-stack --skill gk-graph
Please help me install this Agent Skill. Skill: gk-graph Source: https://github.com/gkganesh12/gk-stack/tree/main/skills-library/gk-graph Command: npx skills add https://github.com/gkganesh12/gk-stack --skill gk-graph
This skill solves the problem of excessive token consumption and context window saturation by providing a lightweight, queryable dependency graph that replaces the need to load full codebase documentation.
Run the gk-graph build command to initialize the dependency and concept graphs for your current project.
High-intent search queries and answers about installing and using this skill.
To reduce token consumption during codebase analysis, you can build a queryable dependency graph that replaces loading full documentation, directly preventing context window saturation.
Blast radius analysis identifies exactly which files are impacted by a code change. It allows AI agents to query a dependency graph, significantly reducing the token budget required for diagnosis and testing.
Extract design rationale by automatically surfacing WHY, HACK, and DECISION comments from source code. This maintains a concept graph that prevents intent drift during bug fixes or code reviews.
No, gk-graph is a zero-dependency skill. It operates entirely using native git and filesystem commands to ensure high-efficiency context management without requiring external packages.
Map both layers by maintaining a dual-layer dependency and concept graph. This tracks file-level technical dependencies alongside concept-level semantic relationships across the project.
The best way is to analyze a dual-layer dependency graph that maps file and semantic relationships. This approach highlights god nodes and prevents architectural context overload.
eliottbusinessAutomate GitHub code reviews with inline comments and PR analysis.
manuelbrandner85Parse source code into directed graphs for security-focused static analysis.
SayBGMManage token budgets and persist session state across multi-agent workflows.
HCMUTE-RTICAnalyze code directories to generate C4 code-level documentation.
