gdgraphlisted

# gdgraph Skill Use this skill by default for project navigation and file discovery. The user does not need to explicitly ask for graph usage. When command output, search results, diff, logs, or large file reads may be long, pair this with `skills/gdctx/SKILL.md` so graph narrows the file set and gdctx compresses the output. Run gdgraph before raw file search when the task involves finding relevant files, understanding project structure, implementation, review, refactoring, debugging, code understanding, impact analysis, architecture, dependencies, or navigation. A "targeted" `rg` does not exempt you: any text, symbol, or pattern search over project code is a search step and goes through the routing layer, not a bare `rg`/`grep`. Skip gdgraph only when the request is clearly unrelated to project files, asks for a single known file's literal contents, or when gdgraph is unavailable. Skipping gdgraph is NOT permission to run raw `rg`: when the graph cannot seed the first hop (unknown symbol, no known file), do the text search with `keryx ctx rg "<pattern>"`, then feed the seed file back into `keryx gdgraph affected <file>`. Raw `rg`/`grep` is a last resort only, and only with a stated reason. ## Trigger Examples - "Добавь обработку ошибки в init." - "Проверь этот модуль." - "Почему этот импорт ломается?" - "Где лучше изменить эту логику?" - "Где лежит логика инициализации?" - "Какие файлы связаны с модулем gdgraph?" - "Найди, где описаны rules/skills." - "Что затронет изм