@imogenlabs/operator-kit

AI operator scaffold for OpenClaw. One install gives you a complete AI operator setup — identity, memory system, instincts engine, cron jobs, and multi-agent architecture.

"Your AI assistant, ready to go."

Install

openclaw plugins install @imogenlabs/operator-kit

What Gets Scaffolded

~/.openclaw/workspace/
├── SOUL.md                    # Core values, boundaries, working style
├── IDENTITY.md                # Name, role, personality
├── USER.md                    # Owner profile and preferences
├── AGENTS.md                  # Memory system, instincts, agent architecture
├── TOOLS.md                   # Services, APIs, MCP servers, models
├── HEARTBEAT.md               # Proactive check schedule
│
├── memory/
│   ├── MEMORY.md              # Index of references and daily logs
│   ├── 2026-03-16.md          # Daily session log (auto-created)
│   └── ref-<topic>.md         # Evergreen references (created as needed)
│
├── instincts/
│   └── global/
│       └── ask-before-external.yaml   # Seed instinct (confidence: 0.9)
│
├── agents/                    # (optional) Multi-agent roles
│   ├── orchestrator/SOUL.md   # Strategy, dispatch, full capabilities
│   ├── coder/SOUL.md          # Code changes, builds, tests
│   ├── security/SOUL.md       # Monitoring, anomaly detection
│   └── researcher/SOUL.md     # External research, fact-checking
│
└── cron/                      # (optional) Starter automation
    └── jobs.json              # Morning briefing + nightly rundown

What You Get

Component	Files	Description
Identity	SOUL.md, IDENTITY.md, USER.md	Core values, personality, owner profile
Workspace	AGENTS.md, TOOLS.md, HEARTBEAT.md	Rules, tools config, proactive checks
Memory	memory/MEMORY.md, memory/YYYY-MM-DD.md	Daily logs + evergreen references
Instincts	instincts/global/*.yaml	Confidence-weighted learned behaviors
Agent Roles	agents/{orchestrator,coder,security,researcher}/	Multi-agent dispatch architecture
Cron Jobs	cron/jobs.json	Morning briefing + nightly rundown

Configuration

{
  "operator-kit": {
    "enabled": true,
    "config": {
      "operatorName": "Friday",
      "ownerName": "Tony",
      "timezone": "America/New_York",
      "enableMemorySystem": true,
      "enableInstincts": true,
      "enableCronJobs": false,
      "enableAgentRoles": false
    }
  }
}

Option	Default	Description
operatorName	"Assistant"	Your operator's name
ownerName	"User"	Your name
timezone	"America/New_York"	IANA timezone for cron scheduling
enableMemorySystem	true	Daily logs + evergreen references
enableInstincts	true	Confidence-weighted learned behaviors
enableCronJobs	false	Morning briefing + nightly rundown
enableAgentRoles	false	Multi-agent architecture

Tools

The plugin registers three tools your operator can use:

Tool	Description
operator_scaffold_status	Check which operator files exist and which are missing
operator_create_instinct	Create a new learned behavior pattern (yaml in instincts/global/)
operator_memory_ref	Create or update an evergreen reference memory file

How It Works

On first load, operator-kit checks your workspace for missing files and creates them from templates. Existing files are never overwritten — you own your operator's identity once it's scaffolded.

The memory system creates daily log files and an index. The instincts engine seeds one starter instinct (ask before external actions) and provides a tool to create more as your operator learns.

Instincts Format

id: prefer-simple-solutions
created: 2026-03-16
confidence: 0.7          # 0.3 weak → 0.9 proven
domain: coding
trigger: "choosing between approaches"
action: |
  Pick the simplest solution that works.
  Three lines of code beats a premature abstraction.
rationale: |
  Over-engineering wastes time and creates maintenance burden.
evidence:
  - date: 2026-03-16
    what: Rewrote 200-line helper as 15-line inline function

Companion: Lossless Context Management

For cross-session memory that survives context window limits, install lossless-claw:

openclaw plugins install @martian-engineering/lossless-claw

Operator-kit handles the identity and workspace structure. Lossless-claw handles the conversation memory — DAG-based summarization with incremental compaction so your operator remembers what happened across sessions.

Inspired By

Built from the patterns behind Imogen, the AI operator at ImogenLabs.

Links

• Product page — features and documentation
• npm — package registry
• OpenClaw — the platform this plugin runs on
• ImogenLabs — who built this

License

MIT