| 1 | // Guards Eleven's knowledge base. node test/knowledge.test.mjs |
| 2 | // This text is injected into Eleven's system prompt and is effectively extractable |
| 3 | // by anyone chatting with it, so the critical invariant is: it contains OPEN |
| 4 | // design knowledge and NO operational secrets. This test fails loudly if a future |
| 5 | // edit ever pastes an address, token, key, or instance name into it. |
| 6 | import assert from "node:assert/strict"; |
| 7 | import { APP_KNOWLEDGE } from "../agent/knowledge.mjs"; |
| 8 | |
| 9 | let failures = 0; |
| 10 | function test(name, fn) { |
| 11 | try { |
| 12 | fn(); |
| 13 | console.log(`PASS ${name}`); |
| 14 | } catch (e) { |
| 15 | failures++; |
| 16 | console.log(`FAIL ${name} — ${e && e.message}`); |
| 17 | } |
| 18 | } |
| 19 | |
| 20 | test("the knowledge base is substantial (not accidentally emptied)", () => { |
| 21 | assert.equal(typeof APP_KNOWLEDGE, "string"); |
| 22 | assert.ok(APP_KNOWLEDGE.length > 10000, `too short: ${APP_KNOWLEDGE.length} chars`); |
| 23 | // Spot-check that it actually covers the core topics. |
| 24 | for (const topic of [/end-to-end/i, /thread/i, /invite/i, /remind/i, /Announcer/, /passkey|backup/i]) { |
| 25 | assert.match(APP_KNOWLEDGE, topic); |
| 26 | } |
| 27 | }); |
| 28 | |
| 29 | test("the knowledge base leaks NO operational secrets", () => { |
| 30 | const forbidden = [ |
| 31 | [/\b\d{1,3}\.\d{1,3}\.\d{1,3}\.\d{1,3}\b/, "an IP address"], |
| 32 | [/run\.garden|elevenmessenger\.com|paulca/i, "a real hostname/instance"], |
| 33 | [/hetzner|\bmiren\b|1password|\bop\b\s+(?:read|item)/i, "an operator/tooling name"], |
| 34 | [/\.p8\b|J5NDF5|ZKYQ9|9X6NKTF|admintok|LCHAT_ADMIN_TOKEN|Bearer\s+\w/i, "a key/token"], |
| 35 | [/BEGIN (?:EC |RSA |OPENSSH )?PRIVATE KEY/i, "a private key"], |
| 36 | [/ssh\s+root@|root@\d/i, "an SSH target"], |
| 37 | ]; |
| 38 | for (const [re, what] of forbidden) { |
| 39 | assert.doesNotMatch(APP_KNOWLEDGE, re, `knowledge base appears to contain ${what}`); |
| 40 | } |
| 41 | }); |
| 42 | |
| 43 | console.log(failures === 0 ? "\nALL PASSED" : `\n${failures} FAILED`); |
| 44 | process.exit(failures === 0 ? 0 : 1); |
| 45 | |
amadan