Maintain project file-tree and code-structure snapshots in a SQLite cache.
npx skills add https://github.com/YuluoY/nimis --skill project-context-yuluoy
Please help me install this Agent Skill. Skill: project-context Source: https://github.com/YuluoY/nimis/tree/main/skills-en/project-context Command: npx skills add https://github.com/YuluoY/nimis --skill project-context-yuluoy
AI coding agents frequently lose awareness of a project's file tree and exported symbols across sessions, forcing repeated rescans and causing missed context; project-context solves this by maintaining a persistent, queryable snapshot of the project's structure and code summaries in a local SQLite cache.
Run the init action to create a project snapshot and persist structure data to .cache/context.db.
High-intent search queries and answers about installing and using this skill.
To persist project structure across sessions, run an incremental sync that records file metadata, mtime, and hash into a local SQLite cache (.cache/context.db), preventing AI agents from losing project context without modifying source files.
Incremental dependency scanning works by using mtime and hash checks to sync file changes, extracting inter-file import graphs and code summaries to detect stale entries and record deleted files in a local SQLite database.
To extract knowledge from code modules, use CLI operations to insert knowledge_edges and knowledge_flows during the Deliver stage, integrating a phase guard for Plan, Execute, Validate, and Deliver agent workflows.
Yes, you can query exported symbols and module files directly from the persistent SQLite context database, using validate and query CLI operations to list metadata and inspect focused modules without triggering a full project rescan.
No, this project context tracking approach does not modify source code files; it performs read-only incremental syncs of file metadata, categories, and dependency graphs, storing all extracted context exclusively in the .cache/context.db database.
You should run a stale-check when AI agents need accurate project awareness, using the CLI to detect stale database entries, mark deleted files, and validate that inter-file dependencies and code summaries match the current file-tree state.
adamrdrewCreate a database directory and empty index.md file idempotently.
lync-cyberRender project metadata, logs, and dependencies into visual summaries.
AskTinNguyenOrchestrate the Work phase by executing plan tasks with agent mindsets and validation.
bahayonghangIndex Markdown notes into a local SQLite vector/full-text index.
