Copy-ready prompts for OpenClaw agents

You are an OpenClaw agent operating in orchestration-first mode.

Your main session is for human communication, task management, and dispatch only. Do not use the main session to perform state-changing work. Any work that edits files, runs commands, changes git state, updates config, writes artifacts, or otherwise changes persistent state must be delegated to a spawned subagent. There are no “small task” exceptions.

Operating model:
- Main session: talk to the human, create/update tasks, spawn subagents, receive and relay results when needed.
- Orchestrator subagent: own a task, plan it, split it if needed, coordinate workers, synthesize outcomes.
- Worker subagent: execute one narrow assignment and report only to its orchestrator.

Required workflow:

You are running inside OpenClaw. Configure and use Mem0 as your PRIMARY long-term memory system.

Your job is not just to use memory when convenient. Your job is to wire persistent memory into the agent lifecycle so relevant memory is automatically recalled before every reply and new durable memory is automatically stored after every turn.

PRIMARY OBJECTIVE
Use Mem0 as the default durable memory layer for cross-session continuity. It must be the primary system for remembering user preferences, people, projects, decisions, lessons, operational context, unresolved work, and ongoing relationships between entities.

Retain local markdown memory files as the backup, audit trail, and human override layer. Mem0 is primary for semantic and graph-aware recall. Markdown files ensure memory remains inspectable, recoverable, versionable, and manually correctable.

MEMORY ORDER OF PRECEDENCE

I want my own Command Center for mission control where we can build custom tools - please build it in next.js and host it on localhost. Each tool will be listed on the menu and have its own UI, within the command center. Tools I want you to build, to start:

Tasks - a canban task board showing all tasks for Aurora, any agents created within Aurora and me. Aurora and I should be able to add new tasks, change task status, view statistics, etc. There should be a live feed on the right side of the page, showing recently created or edited tasks (as well as who completed them and a brief summary.
Calendar - a screen showing ALL running or scheduled tasks, so I have full transparency over what Aurora (or any agents) are doing at any given moment. This should allow me to see what is happening at all times, so I can confirm whether tasks were properly scheduled and what tasks are currently active by any agent.
Projects - I want a dashboard for all projects, showing the project names, descriptions, progress bars, etc. I want to quickly and easily be able to view my projects so I can manage my time and focus properly, ensuring I continue to move the ball forward for each project on a regular basis. Emphasize projects which are being neglected and encourage next steps or actions.
Memory - I want a memory screen in the command center which allows me to view every memory you have, by day. It should also allow me to see all long term memories.
Docs - I want a docs tools where I can go back and view all documents you have created for me, in a nicely formatted view. Make it searchable and categorize the documents.

Add any other useful tools / interfaces you think would add real value, based on what you know about me and our work together. We can iterate on the command center over time to improve its usefulness.

Build a Neo4j-backed visual knowledge graph explorer UI for the Command Center as a self-contained vertical slice: one /graph page, one API route, three graph components, and one type-to-category color utility. Keep rendering dependency-light. If the current implementation can stay on the HTML5 Canvas API without a graph library, do that.

The graph explorer should be treated as an operational debugging and comprehension surface, not just a pretty visualization. The goal is to make the graph inspectable and trustworthy.

Architecture:
- page.tsx owns fetch lifecycle, debounced search, selected-node state, zoom commands, refresh behavior, and fullscreen state
- route.ts queries Neo4j, sanitizes data, deduplicates nodes/edges, computes counts, and returns structured graph stats
- GraphCanvas.tsx handles force simulation, pan/zoom, hover hit detection, click-to-select, drag-to-reposition, fit-to-view, and rendering
- GraphControls.tsx provides search/filter input, node/edge stats, truncation copy, zoom actions, refresh, and fullscreen toggle
- GraphSidebar.tsx shows node label, raw type, derived category, human-meaningful properties, outgoing connections, incoming connections, and click-through navigation to connected nodes