Where is the graph data stored?

All knowledge is stored locally in your project directory at `.agent/knowledge/graph.json`, allowing it to be version-controlled with your code.

Does it automatically capture my corrections?

Yes, when integrated with nav-profile, it can automatically convert user corrections into persistent graph memories to ensure Claude doesn't repeat the same error twice.

How does the Knowledge Graph help with long sessions?

It implements context engineering by storing knowledge in a graph. Instead of Claude reading your entire history, this skill only loads relevant nodes (1-2k tokens) based on your current query, preventing context crashes.

What types of memories can I store?

You can store four specific types: Patterns (conventions), Pitfalls (common mistakes), Decisions (architecture choices), and Learnings (domain insights).

Navigator Knowledge Graph

Name: Navigator Knowledge Graph
Author: alekspetrov

byalekspetrov

•

147

协作工具

Manages a persistent project knowledge graph to capture architectural decisions, pitfalls, and task context for long-running AI coding sessions.

关于

The Navigator Knowledge Graph skill transforms Claude into a context-aware development partner by maintaining a structured map of your project's history, SOPs, and experiential learnings. By indexing tasks, decisions, and 'pitfalls' in a local graph, it allows Claude to surface only the most relevant information for your current task, preventing context window bloat and ensuring critical project nuances are never forgotten across sessions. This skill is essential for complex projects where maintaining architectural consistency and remembering previous debugging insights is critical.

主要功能

Context-aware retrieval that optimizes token usage by loading relevant nodes
Unified search across tasks, SOPs, and system documentation
Experiential memory capture for project-specific patterns and pitfalls
Relationship traversal to find connected concepts and documents
Automatic synchronization of user corrections into persistent memories
147 GitHub stars

使用场景

Querying all project knowledge and previous decisions related to a specific module like 'authentication'
Capturing a 'pitfall' memory after a difficult bug fix to ensure the same mistake isn't repeated
Initializing a new project graph from existing documentation to onboard Claude to a codebase

关于

主要功能

Context-aware retrieval that optimizes token usage by loading relevant nodes
Unified search across tasks, SOPs, and system documentation
Experiential memory capture for project-specific patterns and pitfalls
Relationship traversal to find connected concepts and documents
Automatic synchronization of user corrections into persistent memories
147 GitHub stars

使用场景

Querying all project knowledge and previous decisions related to a specific module like 'authentication'
Capturing a 'pitfall' memory after a difficult bug fix to ensure the same mistake isn't repeated
Initializing a new project graph from existing documentation to onboard Claude to a codebase