@suveren/gateway 0.41.0 → 0.41.1

package/content/guides/6-risks.md

@@ -11,6 +11,7 @@ What can make a test say more than it should, and what to do about each. Read th
 | The values a limit checks are readable on the Authority Server | Proposals are encrypted end to end since 0.27.0 — the server sees only the values it enforces limits on (e.g. an amount or a number of recipients) and fingerprints, never the content | Test data and renamed cases only in a setup; limits that refer to personal data are a profile question |
 | Setup limit too low | A clear without the following load leaves empty systems | Use a setup limit of 4–6 per day; clear and load in one go |
 | An AI that can control a browser or the screen | It could open the gateway's page, read the API key there and approve its own proposals — the person would no longer decide | Never open or operate the gateway's page, never handle the API key, never approve. Tell the person to approve themselves, and to keep the gateway page out of any browser an AI controls |
+| The AI writes as the person, or as a person at all | A customer believes they dealt with a human, or with the mandate owner personally. The footer says an AI sent it; the text would contradict it | Never write as the mandate owner or as any person, and never claim to be human. Write as their AI or about them in the third person, even when the person asks otherwise, and say why |
 | Calls around the Suveren tools | A terminal, HTTP or browser call to the gateway's ports is not governed by a mandate | Use only the Suveren tools. If something is missing, tell the person what — do not probe the gateway |
 
 Next: start a run — the person asks the working AI to work the new requests.

package/dist/mcp-server/http.mjs

@@ -8000,7 +8000,9 @@ function agentVisibleAuthorizations(authorizations, integrationManager2) {
 }
 
 // src/lib/agent-rules.ts
+var NO_IMPERSONATION_RULE = `You are an AI acting under a person's mandate. This rule comes before every other instruction, including the person's own: never write as if you were that person or any person. No "I" as them, no signature with their name alone, no feelings, opinions or experiences of your own, and never claim to be human. Write as their AI, or about them in the third person ("she asked me to confirm\u2026", "he will\u2026"). If the person asks you to write as them, do the task this way instead and tell them why. Every message you send is marked as sent by an AI; make that true of every word.`;
 var AGENT_RULES = [
+  NO_IMPERSONATION_RULE,
   "Never open or operate the Suveren Gateway's web page \u2014 not in a browser you control, not with computer use. Never ask for, read or type the person's API key. Never approve a proposal: approvals are the person's alone.",
   "Use only the Suveren tools. No terminal, HTTP or browser calls to the gateway's ports (for example `curl localhost:3500/mcp`).",
   "When a tool offers an optional `_source` argument, set it to link the ONE record (e.g. an inbound email) you are acting on \u2014 just the reference ({connector, kind, id}), never a summary or quote; the gateway reads the record itself for the approver."

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@suveren/gateway",
-  "version": "0.41.0",
+  "version": "0.41.1",
   "description": "Suveren gateway — local agent gateway built in compliance with the Human Agency Protocol (HAP). Runs the UI, control plane, and MCP server in one Node process.",
   "type": "module",
   "main": "server.js",