gh-macroman5-claude-code-workflow-plugins/.claude/commands/memory-graph.md

description, argument-hint, allowed-tools

description	argument-hint	allowed-tools
Manage persistent knowledge via MCP Memory Graph	intent	Read, Write, Grep, Glob, Bash, Task

Memory Graph Command

This command activates the Memory Graph Skill and guides you to use the MCP Memory server tools to persist, update, search, and prune knowledge.

Inputs

INTENT: ${1:-auto}

Skill

Include the Memory Graph Skill files:

• .claude/skills/memory-graph/SKILL.md
• .claude/skills/memory-graph/operations.md
• .claude/skills/memory-graph/playbooks.md
• .claude/skills/memory-graph/examples.md

If any file is missing, read the repo to locate them under .claude/skills/memory-graph/.

Behavior

• Detect entities, relations, and observations in the current context
• Use MCP tool names prefixed with mcp__memory__
• Avoid duplicates by searching before creating
• Keep observations small and factual, include dates when relevant
• Verify writes with open_nodes when helpful

Planner

1. If INTENT == auto, infer one of: persist-new, enrich, link, correct, prune, explore
2. Route per playbooks.md
3. Execute the needed MCP tool calls
4. Print a short summary of what changed
5. When appropriate, suggest next relations or entities

MCP Tooling

Target server name: memory (tools appear as mcp__memory__<tool>)

Core tools:

• create_entities, add_observations, create_relations
• delete_entities, delete_observations, delete_relations
• read_graph, search_nodes, open_nodes

Examples

Persist a new service and owner

/lazy memory-graph "persist-new service:alpha (owner: alice, repo: org/alpha)"

Explore existing graph

/lazy memory-graph explore

Correct a stale fact

/lazy memory-graph "correct owner for service:alpha -> bob"