Lobster Empire Bootstrap

Use this skill when the goal is not just to copy files, but to turn a new lobster into an Empire-compatible execution unit.

Core outcome

A fresh agent/workspace should gain these capabilities:

• shared execution state via StateDB
• watchdog-based timeout / silence supervision
• lesson sharing across roles
• startup context injection
• Cortex-style coordination discipline
• main-agent DB-first execution habit

Trigger conditions

Use this skill when the user asks to:

• 把龙虾帝国核心架构做成技能
• 初始化一个新龙虾 / 新分身 / 新工作区
• 让一个新 agent 接入帝国
• 复制 / 迁移 / 重建 StateDB + Watchdog + LessonSync 架构
• 给新环境补上帝国执行协议
• bootstrap / transplant / inherit the Lobster Empire architecture

What to read first

Always read these references in this order:

1. references/empire-overview.md
2. references/bootstrap-checklist.md
3. If the task includes real implementation, also read:
• references/file-map.md
• references/adoption-protocol.md

Built-in scripts

Use these when real execution is needed:

• scripts/bootstrap_empire.py — scaffold/copy the minimum Empire stack into a target workspace
• scripts/validate_empire_bootstrap.py — validate whether the target already satisfies the Empire minimum contract

Preferred execution on this machine:

& "C:\Users\cloud_user\AppData\Local\Programs\Python\Python314\python.exe" skills\lobster-empire-bootstrap\scripts\bootstrap_empire.py --target <workspace>
& "C:\Users\cloud_user\AppData\Local\Programs\Python\Python314\python.exe" skills\lobster-empire-bootstrap\scripts\validate_empire_bootstrap.py --target <workspace>

Empire minimum contract

A lobster is considered "joined the Empire" only when all of these are true:

1. Has a local StateDB (state/system.db) with the 5 core tables
2. Has state/task_api.py
3. Has state/context_injector.py
4. Has state/watchdog.py
5. Startup flow reads ContextInjector before task routing
6. Main agent follows DB-first task execution for substantial tasks
7. Lessons are shared cross-role instead of trapped in one agent
8. A watchdog launch path exists (manual or auto-start)

Workflow

1. Identify the target environment:
• same machine new workspace
• another OpenClaw workspace
• partial repair of an existing Empire node
2. Read the architecture references.
3. Inspect whether the target already has any of these:
• state/
• memlib/
• AGENTS.md startup gate
• existing watchdog / task DB
4. Choose one mode:
• bootstrap-new: create the minimum Empire stack from scratch
• adopt-existing: graft Empire protocols onto an existing workspace
• repair-drift: fix a broken or half-installed Empire node
5. Apply files in dependency order:
• state layer
• memlib protocol layer
• startup binding layer
• watchdog launch / health validation
6. Verify with the checklist in references/bootstrap-checklist.md.
7. Report what was inherited, what was skipped, and what still needs manual action.

Required file groups

1) State layer

Mandatory core files:

• state/task_api.py
• state/context_injector.py
• state/watchdog.py
• state/system.db or an init path that creates it

2) Protocol layer

Mandatory mechanism files:

• memlib/state-db-watchdog-lesson-system.md
• memlib/main-agent-statedb-execution-protocol.md
• memlib/cortex-core-interface.md

3) Startup binding layer

At minimum, the target AGENTS.md startup must enforce:

• StateDB/context check before routing
• Cortex core interface read before substantive work

Adoption rules

• Do not claim Empire bootstrap is complete if only files were copied but startup binding was not added.
• Do not claim success if Watchdog exists but cannot be launched.
• Do not leave the target in a "half-Empire" state without explicitly marking missing layers.
• If asked to produce a reusable skill, prefer a concise SKILL.md plus references over a giant monolith.

When to create scripts

Create helper scripts only if the same bootstrap logic will be repeated or if deterministic setup is needed.

Good candidates:

• file copy / scaffold script
• validation script
• watchdog registration helper

This skill already includes:

• scripts/bootstrap_empire.py
• scripts/validate_empire_bootstrap.py

Output expectations

When you finish, provide:

• installed layers
• skipped / blocked layers
• exact files added or expected
• validation result against the minimum contract
• next manual step if any

Guardrails

• Do not silently overwrite a user's existing architecture without telling them.
• For system-critical config or gateway changes, ask first.
• If bootstrap is cross-machine and you cannot access the target, produce a transplant packet instead of pretending it is live.

Success condition

The skill is successful when a future fresh lobster can use this skill to reliably become an Empire-compatible node rather than merely receiving a pile of copied files.