agentdom.vercel.app WebMCP audit

Name: agentdom.vercel.app WebMCP audit
Creator: ora

Rescan

audited Sep 1, 2026in 15s

51/ 100100% of the applicable check weight evaluated

8 tools registered. Strongest in shared experience, weakest in trust.

Shared experience100

Task completion0

Tool quality86

Trust0

https://agentdom.vercel.app

Tools

0 read, 0 write, 8 undeclared

Loading map

list_notes

List the user's local notes. Returns only titles and a short 1-line preview for each note — never the full note body.No behaviour hints declared. Agents should treat it as possibly state-changing.

pagehttps://agentdom.vercel.app/

implementation

viaimperative

entry pointnavigator

registered after183ms

executepresent

api surfaceTakes no parameters.

annotations

read onlynot declared

untrusted contentnot declared

titlenot declared

3 findings

warningCanonical entry point

weight 3

Every tool registers through the deprecated navigator.modelContext alias; move them to document.modelContext before the alias is removed.

fix

// document.modelContext is the canonical entry point - the navigator.modelContext alias is deprecated.
document.modelContext.registerTool({ /* ... */ });

failAnnotations present

weight 6

4 of 4 read-shaped tools do not declare readOnlyHint: true, so an agent has to treat them as possible writes and ask before calling. (Declaring readOnlyHint: false earns nothing - it is the default.)

list_notes, search_notes, summarize_selection, suggest_links

fix

{
  name: "list_notes",
  // Declare readOnlyHint: true on tools that truly have no side effects -
  // that is the claim that lets an agent relax confirmation on reads.
  // (false is the default, so declaring it adds no information.)
  annotations: { readOnlyHint: true }
}

warningUntrusted content hint

weight 4

2 of 2 tools look like they return text written by other people but declare no untrustedContentHint, so an agent will treat the response as the site speaking.

list_notes: note, user; suggest_links: note, user

fix

{ name: "list_notes", annotations: { untrustedContentHint: true } }

tool json

{
  "name": "list_notes",
  "description": "List the user's local notes. Returns only titles and a short 1-line preview for each note — never the full note body.",
  "inputSchema": {
    "type": "object",
    "properties": {},
    "additionalProperties": false
  },
  "annotations": {}
}

Findings

Shared experience

100 / 100 · weight 30

passPage experience

weight 20

The page a person sees holds up next to the agent surface: a working interface, visible actions, readable content, nothing in the way.

passHuman parity

weight 10

The person co-browsing can see and use this page - the same page the agent's tools act on.

Task completion

0 / 100 · weight 25

failTool selection

weight 15

An agent chose an existing tool and built a schema-valid call that held up on 0 of 3 canonical intents; on "Open the document I was working on most recently." no tool on the page served it. Nothing was executed, so this verifies selection, not outcomes.

failCoverage vs. site type

weight 10

The tool set covers 0 of the 3 things an agent needs on a editor site; it cannot open or read a document or edit or write content or save, publish, or share the result.

Tool quality

86 / 100 · weight 25

passRegistration timing

weight 1

Every measured tool registered within 183ms of navigation.

toolsuggest_links

warningCanonical entry point

weight 3

Every tool registers through the deprecated navigator.modelContext alias; move them to document.modelContext before the alias is removed.

toollist_notes

fix

// document.modelContext is the canonical entry point - the navigator.modelContext alias is deprecated.
document.modelContext.registerTool({ /* ... */ });

passSchema validity

weight 4

All 8 declared input schemas are structurally valid object schemas.

warningSchema quality

weight 4

2 of 8 tool schemas are harder for an agent to use than they need to be: accept_suggestion 1 parameter has no description (suggestionId).

toolaccept_suggestion

accept_suggestion: 1 parameter has no description (suggestionId); reject_suggestion: 1 parameter has no description (suggestionId)

fix

{
  name: "accept_suggestion",
  inputSchema: {
    type: "object",
    properties: {
      query: { type: "string", description: "What to search for" }
    },
    required: ["query"]
  }
}

warningNaming quality

weight 2

Tool naming makes selection harder than it needs to be: 3 names do not start with a verb (highlight_related).

toolhighlight_related

fix

{ name: "highlight_related" /* short, unique, verb-based */ }

passStub detection

weight 4

All 8 tools declare an execute handler.

passRegistration errors

weight 2

No tool registration threw during the capture.

passDescription quality

weight 5

All 8 rated tool descriptions say what the tool does, when to use it, and what it returns.

Trust

0 / 100 · weight 20

failAnnotations present

weight 6

4 of 4 read-shaped tools do not declare readOnlyHint: true, so an agent has to treat them as possible writes and ask before calling. (Declaring readOnlyHint: false earns nothing - it is the default.)

toollist_notes

list_notes, search_notes, summarize_selection, suggest_links

fix

{
  name: "list_notes",
  // Declare readOnlyHint: true on tools that truly have no side effects -
  // that is the claim that lets an agent relax confirmation on reads.
  // (false is the default, so declaring it adds no information.)
  annotations: { readOnlyHint: true }
}

not applicableAnnotation mismatch

weight 6

No tool declares a readOnlyHint, so there is no safety claim to contradict.

warningUntrusted content hint

weight 4

2 of 2 tools look like they return text written by other people but declare no untrustedContentHint, so an agent will treat the response as the site speaking.

toollist_notes

list_notes: note, user; suggest_links: note, user

fix

{ name: "list_notes", annotations: { untrustedContentHint: true } }

not applicableInjection surface

weight 4

No tool's metadata carries instruction-shaped text aimed at the agent reading it. This check only ever flags; it never credits a site.

Tool selection

0% across 3 intents

Open the document I was working on most recently.missed

choseno toolexpected document openkind not applicable

No tool was picked for this intent.

arguments

{}

Add a short summary paragraph at the top.missed

choseno toolexpected content editkind not applicable

No tool was picked for this intent.

arguments

{}

Save it and give me a link I can share.missed

chosecreate_safe_summaryexpected save or sharekind mismatch

The tool the model chose does not look like the kind of tool this intent needs.

arguments

{
  "noteId": ""
}

model: gemini-2.5-flash

Add the tools this site is missing

Our scanner reads your website and suggests the right WebMCP tools for it.

The open source webmcp plugin teaches your coding agent to audit a site, implement tools on document.modelContext, and verify them in a real browser. npx @ora-ai/webmcp-verify runs the verification on its own. No signup, no hosted service.

how this was captured

observed via: capture shim (Chromium 148.0.7778.96)
chrome: 148.0.7778.96
capture shim: v1
spec snapshot: 2026-08-26
mode: fast
pages: 1 - entry page only
tools: 8
run time: 15s
finished: Sep 1, 2026

Checked the way in-browser agents discover tools: the top-level document's modelContext registry, read after the page settles.