Skip to content

Open-source platform for 3D AI agents. Turn text or a photo into a rigged, animated GLB avatar, give it an LLM brain, memory and a wallet, and embed it anywhere with one web component. Remote MCP server, x402 payments, Three.js, browser-native.

docs(guards): register the two new MCP guards so the guard page lists them

nirholas26fe26e3dc68
diff --git a/data/guards.json b/data/guards.json
index cad593060..cee437f47 100644
--- a/data/guards.json
+++ b/data/guards.json
@@ -1,5 +1,5 @@
{
- "updated": "2026-08-06",
+ "updated": "2026-09-03",
"summary": "Every mechanical guard that protects this repository, and the stage each one runs in. This file is the single source of truth: scripts/audit-guards.mjs proves every entry is real and every stage claim is true, docs/guards.md explains them, and /guards renders them.",
"stages": [
{
@@ -412,6 +412,43 @@
"expect": "public/mcp-catalog.json is stale"
}
},
+ {
+ "id": "audit-mcp-listing",
+ "script": "scripts/build-mcp-listing-source.mjs",
+ "npm": "audit:mcp-listing",
+ "title": "MCP directory listing source",
+ "protects": "prompts/store-submissions/_generated/mcp-listing-source.json describes every MCP server we publish, at the version we publish.",
+ "why": "Every third-party directory listing (the official registry, Smithery, Glama, mcp.so, PulseMCP, LobeHub) is written from that one file. While it was kept by hand it drifted six servers short and thirty-five versions behind, and nothing noticed, because a listing nobody reads back looks identical whether it is right or wrong. The generator reads the facts from the manifests and this check fails when the committed file no longer matches them, or when a new server lands with no listing copy at all.",
+ "stages": [
+ "gate"
+ ],
+ "needs": "none",
+ "proof": {
+ "summary": "A new MCP server landing with no listing copy, so its directory card would be blank.",
+ "violation": {
+ "write": {
+ "packages/guard-proof-mcp/server.json": "{\n\t\"$schema\": \"https://static.modelcontextprotocol.io/schemas/2025-12-11/server.schema.json\",\n\t\"name\": \"io.github.nirholas/guard-proof-mcp\",\n\t\"title\": \"Guard proof server\",\n\t\"description\": \"Proof fixture server used only inside the guard sandbox.\",\n\t\"websiteUrl\": \"https://three.ws\",\n\t\"repository\": {\n\t\t\"url\": \"https://github.com/nirholas/three.ws\",\n\t\t\"source\": \"github\"\n\t},\n\t\"version\": \"0.0.1\",\n\t\"packages\": [\n\t\t{\n\t\t\t\"registryType\": \"npm\",\n\t\t\t\"identifier\": \"@three-ws/guard-proof-mcp\",\n\t\t\t\"version\": \"0.0.1\",\n\t\t\t\"runtimeHint\": \"npx\",\n\t\t\t\"transport\": {\n\t\t\t\t\"type\": \"stdio\"\n\t\t\t}\n\t\t}\n\t]\n}\n"
+ }
+ },
+ "expect": "no curated listing copy"
+ }
+ },
+ {
+ "id": "audit-mcp-reviewer",
+ "script": "scripts/audit-mcp-reviewer-path.mjs",
+ "npm": "audit:mcp-reviewer",
+ "title": "Connector reviewer path",
+ "protects": "Every paid tool on the published stdio server answers an unpaid call with a clean, priced x402 challenge, and the reviewer entitlement lifts it only for the right secret.",
+ "why": "A Claude Connectors reviewer drives the server with no wallet, so what they see for a paid tool IS the product. A crash, a stack trace or an unexplained JSON blob reads as a broken connector and costs a submission cycle. This drove the whole surface as a reviewer's host does and immediately found the standing evidence was four tools out of date.",
+ "stages": [
+ "manual"
+ ],
+ "needs": "network",
+ "proof": {
+ "kind": "live",
+ "reason": "It spawns the real published server over stdio and calls every tool, and several tools reach live upstreams. A static fixture cannot produce a payment challenge from the x402 wrapper, which is the thing under test."
+ }
+ },
{
"id": "audit-x402-catalog",
"script": "scripts/audit-x402-catalog.mjs",
diff --git a/docs/guards.md b/docs/guards.md
index e9420d057..7d8069428 100644
--- a/docs/guards.md
+++ b/docs/guards.md
@@ -125,6 +125,7 @@ Run `npm run audit:guards` to print the current count and per-stage breakdown. T
| MCP golden contracts | `npm run audit:mcp-golden` | Tool names, descriptions, and schemas against a committed snapshot. |
| MCP safety annotations | `npm run audit:mcp-safety` | Declared `readOnlyHint` and `destructiveHint` match what handlers do. |
| MCP tool catalog freshness | `npm run audit:mcp-catalog` | `public/mcp-catalog.json` matches what the MCP servers actually expose. |
+| MCP directory listing source | `npm run audit:mcp-listing` | The one file every third-party MCP directory listing is written from still matches the manifests, and no new server lands without listing copy. |
| 3D Studio OpenAPI sync | `npm run check:studio-openapi` | The Actions file in the OpenAI submission kit is byte-identical to the OpenAPI schema the site serves. |
| Live event window | `npm run check:event` | `public/event.json` describes an event that will actually happen, on every surface that reads it. |
| Cron schedule drift | `npm run check:cron-syntax`, `npm run check:cron-drift` | Valid expressions, and agreement with the running Cloud Scheduler jobs. |
@@ -160,6 +161,7 @@ Run `npm run audit:guards` to print the current count and per-stage breakdown. T
| Home credential health | `npm run audit:home-credentials` | Every stored Home Assistant token still decrypts under the current `WALLET_ENCRYPTION_KEY`, so no connected house is sealed. |
| Hands-free voice loop | `npm run check:home-voice` | Nothing about listening loads before opt-in, an ambient "yeah" never confirms a guarded action, and the agent does not wake itself. |
| Delegation contract addresses | `npm run check:erc7710` | Every delegation-manager address is a deployed contract. |
+| Connector reviewer path | `npm run audit:mcp-reviewer` | Every paid tool on the published stdio server answers an unpaid call with a clean, priced x402 challenge, and the reviewer entitlement lifts the paywall only for the right secret. |
---
diff --git a/public/changelog.json b/public/changelog.json
index 8611f2cb6..547c41219 100644
--- a/public/changelog.json
+++ b/public/changelog.json
@@ -1,5 +1,5 @@
{
- "generated_at": "2026-09-03T08:05:25.752Z",
+ "generated_at": "2026-09-03T08:11:22.831Z",
"generated_by": "scripts/build-page-index.mjs from data/pages.json + data/changelog.json",
"site": {
"name": "three.ws",
diff --git a/public/guards.json b/public/guards.json
index cad593060..cee437f47 100644
--- a/public/guards.json
+++ b/public/guards.json
@@ -1,5 +1,5 @@
{
- "updated": "2026-08-06",
+ "updated": "2026-09-03",
"summary": "Every mechanical guard that protects this repository, and the stage each one runs in. This file is the single source of truth: scripts/audit-guards.mjs proves every entry is real and every stage claim is true, docs/guards.md explains them, and /guards renders them.",
"stages": [
{
@@ -412,6 +412,43 @@
"expect": "public/mcp-catalog.json is stale"
}
},
+ {
+ "id": "audit-mcp-listing",
+ "script": "scripts/build-mcp-listing-source.mjs",
+ "npm": "audit:mcp-listing",
+ "title": "MCP directory listing source",
+ "protects": "prompts/store-submissions/_generated/mcp-listing-source.json describes every MCP server we publish, at the version we publish.",
+ "why": "Every third-party directory listing (the official registry, Smithery, Glama, mcp.so, PulseMCP, LobeHub) is written from that one file. While it was kept by hand it drifted six servers short and thirty-five versions behind, and nothing noticed, because a listing nobody reads back looks identical whether it is right or wrong. The generator reads the facts from the manifests and this check fails when the committed file no longer matches them, or when a new server lands with no listing copy at all.",
+ "stages": [
+ "gate"
+ ],
+ "needs": "none",
+ "proof": {
+ "summary": "A new MCP server landing with no listing copy, so its directory card would be blank.",
+ "violation": {
+ "write": {
+ "packages/guard-proof-mcp/server.json": "{\n\t\"$schema\": \"https://static.modelcontextprotocol.io/schemas/2025-12-11/server.schema.json\",\n\t\"name\": \"io.github.nirholas/guard-proof-mcp\",\n\t\"title\": \"Guard proof server\",\n\t\"description\": \"Proof fixture server used only inside the guard sandbox.\",\n\t\"websiteUrl\": \"https://three.ws\",\n\t\"repository\": {\n\t\t\"url\": \"https://github.com/nirholas/three.ws\",\n\t\t\"source\": \"github\"\n\t},\n\t\"version\": \"0.0.1\",\n\t\"packages\": [\n\t\t{\n\t\t\t\"registryType\": \"npm\",\n\t\t\t\"identifier\": \"@three-ws/guard-proof-mcp\",\n\t\t\t\"version\": \"0.0.1\",\n\t\t\t\"runtimeHint\": \"npx\",\n\t\t\t\"transport\": {\n\t\t\t\t\"type\": \"stdio\"\n\t\t\t}\n\t\t}\n\t]\n}\n"
+ }
+ },
+ "expect": "no curated listing copy"
+ }
+ },
+ {
+ "id": "audit-mcp-reviewer",
+ "script": "scripts/audit-mcp-reviewer-path.mjs",
+ "npm": "audit:mcp-reviewer",
+ "title": "Connector reviewer path",
+ "protects": "Every paid tool on the published stdio server answers an unpaid call with a clean, priced x402 challenge, and the reviewer entitlement lifts it only for the right secret.",
+ "why": "A Claude Connectors reviewer drives the server with no wallet, so what they see for a paid tool IS the product. A crash, a stack trace or an unexplained JSON blob reads as a broken connector and costs a submission cycle. This drove the whole surface as a reviewer's host does and immediately found the standing evidence was four tools out of date.",
+ "stages": [
+ "manual"
+ ],
+ "needs": "network",
+ "proof": {
+ "kind": "live",
+ "reason": "It spawns the real published server over stdio and calls every tool, and several tools reach live upstreams. A static fixture cannot produce a payment challenge from the x402 wrapper, which is the thing under test."
+ }
+ },
{
"id": "audit-x402-catalog",
"script": "scripts/audit-x402-catalog.mjs",