Skip to content

stunt-double/stuntdouble

v1.14.0MIT

Authentic user insights from autonomous AI agents, backed by evidence.

What this package declares

The file a client reads when it loads this plugin, exactly as this revision carries it.

plugin.json
{
  "$schema": "https://agent-plugins.org/schemas/1.0.0/plugin.schema.json",
  "name": "stuntdouble",
  "version": "1.14.0",
  "description": "Authentic user insights from autonomous AI agents, backed by evidence.",
  "author": {
    "name": "Stunt Double",
    "email": "support@stuntdouble.io",
    "url": "https://www.stuntdouble.io"
  },
  "homepage": "https://www.stuntdouble.io/support/docs/api/mcp",
  "repository": "https://github.com/stunt-double/stuntdouble-mcp",
  "license": "MIT",
  "keywords": [
    "ux-testing",
    "qa",
    "user-research",
    "agent-readiness",
    "accessibility"
  ],
  "extensions": {
    "com.openai": {
      "interface": {
        "displayName": "Stunt Double",
        "shortDescription": "See where people get stuck",
        "longDescription": "Stunt Double gives product teams authentic user insights from autonomous AI agents, backed by evidence. Teams cast AI users, called actors, to match their own audience: each has a background, a goal, a device, a locale and a personal inbox. Given a task, an actor drives a real browser through a product, prototype or Figma design on its own, signing up, verifying its email and choosing every click itself. Every finding comes back with its evidence: the screenshot, the step and what the actor said.\n\nFrom here you can:\n- Read a checklist run as a card: the verdict, each check with the screenshot behind it, and what changed since the last run. The card updates while the run is still going.\n- Run a checklist or automation against production, staging or a preview link.\n- Plan and launch an interview with AI participants, then read the synthesized report.\n- See the themes in what stakeholders said about a project, and triage the feedback.\n- Check how well AI agents can find, understand and act on a website, from the public Stunt Double Index.\n- Search a workspace for the projects, actors, checklists and interviews that already exist.\n\nBuilt for product designers, engineers and QA leads who want to see a flow tried before real people try it.\n\nLimits: the plugin needs a Stunt Double account and reaches only that account's workspaces. Runs take minutes and count toward your plan. Actors reach any public URL and handle email sign-in through their own inboxes; they ask a person only for what a person holds, such as a code sent to a phone. Workspaces, billing and account deletion stay in the Stunt Double dashboard.",
        "developerName": "Stunt Double",
        "category": "Developer Tools",
        "capabilities": [
          "Read checklist runs with per-check evidence",
          "Run checklists and automations against any public URL",
          "Create actors (AI user personas) and checklists",
          "Launch user interviews with AI participants",
          "Summarize and triage stakeholder feedback",
          "Score how agent-ready a website is with the Stunt Double Index",
          "Search a workspace"
        ],
        "websiteURL": "https://www.stuntdouble.io",
        "supportURL": "https://www.stuntdouble.io/support",
        "privacyPolicyURL": "https://www.stuntdouble.io/privacy",
        "termsOfServiceURL": "https://www.stuntdouble.io/terms",
        "defaultPrompt": [
          "Show me the latest run of my checkout checklist and what failed.",
          "How agent-ready is stripe.com, and where do agents get stuck?",
          "What are stakeholders saying in feedback on my project?"
        ],
        "brandColor": "#3555E5",
        "brandColorDark": "#7E8CFC",
        "composerIcon": "./assets/logo.png",
        "composerIconDark": "./assets/logo-dark.png",
        "logo": "./assets/logo.png",
        "logoDark": "./assets/logo-dark.png"
      },
      "review": {
        "test_cases": {
          "positive": [
            {
              "description": "Read a checklist run as a card",
              "prompt": "Show me the latest run of the Checkout checklist in my Demo Shop project and tell me what failed.",
              "tools_triggered": "list_workspaces, list_checklists, get_checklist_run",
              "expected_behavior": "Shows the run card (verdict, score, each check with the actor's evidence, changes since the previous run) and names the failing checks in one or two sentences."
            },
            {
              "description": "Score a public site on the Index",
              "prompt": "How agent-ready is stripe.com, and where do agents get stuck?",
              "tools_triggered": "get_index_report",
              "expected_behavior": "Shows the Index card for stripe.com (score out of 100, band, rank, category scores) and lists the top frictions agents hit, with a link to the public report."
            },
            {
              "description": "Run a checklist and follow it",
              "prompt": "Run the Checkout checklist in Demo Shop again and tell me whether it passes now.",
              "tools_triggered": "list_workspaces, list_checklists, run_checklist, get_checklist_run",
              "expected_behavior": "Starts a run, shows the run card while the actor works (it refreshes itself), and reports the verdict once the run completes."
            },
            {
              "description": "Summarize stakeholder feedback",
              "prompt": "What are people saying in feedback on the Demo Shop project?",
              "tools_triggered": "list_workspaces, list_projects, summarise_feedback",
              "expected_behavior": "Returns the main themes in the project's feedback comments with how many comments sit under each, and an implementation brief."
            },
            {
              "description": "Find existing work before creating more",
              "prompt": "Do I already have anything in Stunt Double that tests checkout?",
              "tools_triggered": "list_workspaces, search",
              "expected_behavior": "Lists the matching checklist, project and actor in the review workspace by name, without creating anything."
            }
          ],
          "negative": [
            {
              "description": "Account deletion is not offered",
              "prompt": "Delete my Stunt Double account and all my data."
            },
            {
              "description": "Billing changes are not offered",
              "prompt": "Upgrade my Stunt Double plan and charge my card."
            },
            {
              "description": "Workspaces outside the connected account",
              "prompt": "Show me the checklists in the Acme Corp workspace."
            }
          ]
        },
        "commerce": false,
        "commerce_description": "Stunt Double does not sell products or process payments through this plugin. Plans are managed in the Stunt Double dashboard."
      },
      "publication": {
        "release_notes": "First release: read checklist runs and Index reports as cards, run checklists and automations, launch AI user interviews, summarize feedback, and search a workspace."
      }
    }
  }
}

Client extensions

Data this package carries for particular clients. The directory lists the clients named and never reads what is addressed to them.

  • com.openai