Use MSG Chain with an AI coding agent让 AI 编码助手接入 MSG 公链
One official prompt plus explicit allowed / forbidden actions, so Cursor, Claude Code, Codex or another agent reads the machine layer instead of guessing.一份官方 prompt 加上明确的允许 / 禁止动作,让 Cursor、Claude Code、Codex 等助手读机器层而不是凭训练数据猜测。
Mainnet not launched (No-Go) · AI has no production validator takeover authority主网未上线(No-Go)· AI 无生产验证者接管权限
Official prompt (copy into your agent)官方 prompt(复制给你的助手)
Give this together with the machine entry URL. It is a boundary contract, not a marketing brief.把这段与机器入口 URL 一起交给助手。它是边界契约,不是宣传稿。
You are integrating with MSG Chain. Follow these rules exactly.
1. Seed from the machine entry:
https://msgchain.org/whitepaper/agent_entry.json
(schema, crawl contract, read order, trust contract)
2. For coding tasks, continue to:
developer_entry.json -> quickstart/ -> recipes/ -> contract_templates/
3. Never upgrade "partial", "planned", or "local" into "ready" or "production".
Every "implemented" claim must bind to an evidence_class of code or onchain_query.
4. Signing, deployment, treasury, and governance actions require human approval.
5. AI has no production validator takeover authority.
6. Answer only from machine-readable evidence. If evidence is missing, say so
and refuse to upgrade the claim (cite-or-refuse).
7. Mainnet status: not launched (No-Go). Read the single status source:
https://msgchain.org/status.json