context-graph

Manually add a new entity to the knowledge graph — prompts for details, creates the entity file, and updates the index

Generate the initial context graph from the current codebase. This is the most important command — it creates the foundation that all other context-graph commands build on.

Search the knowledge graph for entities relevant to a query — searches by name, tags, type, file path, or relation and returns focused context

Generate a visual Mermaid diagram of the context graph.

Audit the context graph for stale, orphaned, or broken entities and optionally fix them.

Use this agent when the user asks to "build the context graph", "create entities", "update the knowledge graph", "add to the graph", "bootstrap the context graph", "document the architecture in the graph", or needs entities created, updated, or maintained in the context graph. <example> User: Build the context graph for this project Use: Yes — full graph bootstrap from codebase analysis <commentary>Routes to /context-bootstrap workflow, analyzes codebase, creates entity files and index.</commentary> </example> <example> User: Add a new entity for the payment service Use: Yes — manual entity creation with relation discovery <commentary>Creates entity file, discovers file associations, establishes relations, updates index.</commentary> </example> <example> User: The auth service now depends on the new cache layer, update the graph Use: Yes — update existing entities with new relations <commentary>Reads both entities, adds depends-on/consumed-by relations, updates Last Updated dates.</commentary> </example> <example> User: Find where the auth module is defined Use: No — this is a read-only search task, use graph-navigator instead <commentary>Navigation and search are graph-navigator's responsibility.</commentary> </example> <example> User: What components depend on the database? Use: No — this is a graph traversal question, use graph-navigator instead <commentary>Read-only graph queries belong to graph-navigator.</commentary> </example>

Use this agent when the user asks to "find in the graph", "what depends on", "show me the architecture", "what entities exist", "how are these related", "navigate the graph", "search the context graph", or needs to explore, query, or understand the knowledge graph without modifying it. <example> User: What components depend on the database module? Use: Yes — graph traversal to find consumers of an entity <commentary>Reads the database entity, follows consumed-by relations, presents dependency chain.</commentary> </example> <example> User: Show me everything related to authentication Use: Yes — tag-based search plus relation traversal <commentary>Greps for "auth" tags, reads matching entities, follows relations for context.</commentary> </example> <example> User: Why did we choose PostgreSQL over MongoDB? Use: Yes — decision audit via the graph <commentary>Finds the relevant decision entity, reads rationale and alternatives considered.</commentary> </example> <example> User: What does the context graph contain right now? Use: Yes — full graph summary from index <commentary>Reads index.md, summarizes entity counts by type, highlights key relationships.</commentary> </example> <example> User: Add a new entity for the payment system Use: No — this is a write operation, use graph-builder or /context-add instead <commentary>Creating or modifying entities belongs to graph-builder.</commentary> </example> <example> User: Update the auth service entity with the new dependency Use: No — this modifies the graph, use graph-builder instead <commentary>Entity updates belong to graph-builder.</commentary> </example>

Maintains a persistent, file-based knowledge graph for project context. Organizes components, decisions, data models, and infrastructure into navigable entities with typed relationships. Unlike session-resilience (ephemeral session state), this is permanent project knowledge that survives across all sessions.

2 hooks across 2 events