Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add instructions/netease-youdao/lobsterai/agents-mdgit clone --depth 1 https://github.com/netease-youdao/LobsterAIWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.04886 | $0.04886 |
| Opus 5 | $0.02443 | $0.02443 |
| Sonnet 5 | $0.00977 | $0.00977 |
| Haiku 4.5 | $0.00489 | $0.00489 |
Grade A, and why
LobsterAI AGENTS.md scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured yesterday.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 524 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md
This file gives coding agents the current working model for this repository.
Treat source code and package.json as the authority when something here
appears stale.
Instruction Scope
This root AGENTS.md is repository-level guidance for LobsterAI. Codex may also
load more specific AGENTS.md or AGENTS.override.md files from subdirectories
when the current working directory is inside those subtrees. More specific
instructions override broader ones.
Do not treat generated runtime workspaces, bundled OpenClaw output, or old Claude memory notes as authoritative project instructions. Use them only as historical context and verify against the current source.
Project Snapshot
LobsterAI is an Electron + React desktop application. Its core user-facing product is a desktop agent experience that can work with local projects, files, browser previews, IM channels, skills, MCP servers, scheduled tasks, and rich artifacts.
Cowork vs OpenClaw
Cowork is LobsterAI's product/session layer. The name is historical: it
started as a Claude Code-like in-house coding assistant, but in the current
codebase it means the LobsterAI layer that owns sessions, messages,
permissions, UI state, local persistence, context usage, artifacts, and IPC
contracts.
OpenClaw is the only agent runtime/gateway. CoworkAgentEngine is currently
'openclaw' only. Legacy names such as cowork:* IPC channels,
claude_session_id, and some "cowork" filenames are compatibility/history
names, not evidence of another active runtime.
Do not reintroduce or design around yd_cowork; it has been removed as a
runtime. If old docs mention it as switchable, verify against current source
before acting.
Commands
# First development run: build/sync the pinned OpenClaw runtime, then start app
npm run electron:dev:openclaw
# Daily development after runtime exists: Vite on port 5175 + Electron
npm run electron:dev
# Production renderer bundle
npm run build
# Electron main/preload TypeScript build
npm run compile:electron
# Full ESLint across src; may fail on existing legacy debt
npm run lint
# CI-equivalent lint for touched TypeScript files
npx eslint --ext ts,tsx --report-unused-disable-directives --max-warnings 0 <files>
# Official test entry used by GitHub CI
npm test
# Filter Vitest tests
npm test -- logger
npm test -- cowork
# Package for distribution
npm run dist:mac
npm run dist:win
npm run dist:linux
# Windows installer gate (runs automatically at the start of dist:win and the
# dist:win:web stub pass): re-applies patches/app-builder-lib+*.patch and runs
# the installer contract tests against node_modules. If it fails, run `npm ci`;
# never ship an installer from a tree where it fails.
npm run verify:installer-patches
# Build current-platform OpenClaw runtime manually
npm run openclaw:runtime:host
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- yesterday First seen · 524 lines · 4,886 tokens per session scan A 19cc0b867c12
LobsterAI AGENTS.md is an instructions file published in the GitHub repository netease-youdao/LobsterAI (5,968 stars, last pushed today), licensed MIT. It adds 4,886 tokens to every session, about $0.0244 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other instructions, from other repositories
hope-agent AGENTS.md
Instructions for shiwenwen/hope-agent, covering hope agent, 安全红线, 提交前检查（强制）, 分支与发布 and 设置约定.
twiko AGENTS.md
Instructions for ezocomp118-source/twiko, a project described as: Self-hosted AI agents with their own computers — browser, terminal, files, and long-term memory. MCP tools, app integrations, and scheduled routines. Bring your own model. Web, desktop, and mobile. Apache-2.0.
hope-agent CLAUDE.md
Instructions for shiwenwen/hope-agent, a project described as: 🦭 会记忆、能持续推进目标、会动态编排多 Agent 的跨端桌面 AI 助手，也可服务化常驻 NAS / 云端 | A cross-device desktop AI agent with memory, autonomous goals, dynamic workflows, and headless deployment.
openagent CLAUDE.md
Claude Code instructions for the-open-agent/openagent, covering claude.md, commands, architecture, backend (go / beego) and frontend (react).
agentsys CLAUDE.md
Claude Code instructions for agent-sh/agentsys, covering agentsys, critical rules, model selection, core priorities and dev commands.
zhin zhin-plugin.instructions.md
Instructions for zhinjs/zhin, covering zhin plugin runtime authoring, package contract, convention directories, imports and native typescript and command routes.