plugin.json

{
  "$schema": "https://agent-plugins.org/schemas/1.0.0/plugin.schema.json",
  "name": "enoki",
  "version": "1.0.0",
  "description": "Test your AI agents for security and safety weaknesses, and re-verify the fixes.",
  "author": {
    "name": "Enoki Labs",
    "email": "hello@enokilabs.ai",
    "url": "https://enokilabs.ai"
  },
  "homepage": "https://docs.enokilabs.ai/connect-your-agent",
  "license": "MIT",
  "keywords": [
    "ai-agents",
    "security",
    "safety",
    "red-teaming",
    "testing"
  ],
  "extensions": {
    "com.openai": {
      "interface": {
        "displayName": "Enoki",
        "shortDescription": "Security testing for AI agents",
        "longDescription": "Enoki tests AI agents for security and safety weaknesses. It attacks your agent the way an adversary would, adapts to how the agent responds, and reports each weakness it finds with the exact attack and the agent's response.\n\nWith Enoki in ChatGPT you can:\n- Register an agent endpoint as a target and keep its profile current.\n- Start an assessment, on everything or on the categories a change touched, and follow it to the end.\n- Read findings with the full attack transcript, so you can fix the weakness in your code or configuration.\n- Close findings, and re-run a finding's attacks against your live agent to confirm a fix worked.\n\nA typical loop: you change your agent's prompt or tools, ask ChatGPT to test the change, read what Enoki found, fix it, and confirm the fix.\n\nYou sign in to Enoki and approve the connection, choosing what ChatGPT may do: read only, change targets and findings, or also start assessments. You can change or revoke that at any time under Settings > Connected apps.\n\nEnoki needs an Enoki workspace; an owner or admin of the workspace approves the connection. Register a test or staging deployment of your agent rather than production, because an assessment sends adversarial messages to it.",
        "developerName": "Enoki Labs",
        "category": "Developer Tools",
        "websiteURL": "https://enokilabs.ai",
        "supportURL": "https://docs.enokilabs.ai/troubleshooting",
        "privacyPolicyURL": "https://enokilabs.ai/privacy",
        "termsOfServiceURL": "https://enokilabs.ai/terms",
        "defaultPrompt": [
          "Test my support agent with Enoki.",
          "What did the last Enoki run find on my agent?",
          "I fixed the finding Enoki reported. Check that it is fixed."
        ],
        "brandColor": "#0F121A",
        "logo": "./assets/logo.png",
        "logoDark": "./assets/logo-dark.png"
      },
      "review": {
        "test_cases": {
          "positive": [
            {
              "description": "List the agents registered in the workspace and when each was last tested.",
              "prompt": "Which agents are registered in my Enoki workspace, and when was each one last tested?",
              "tools_triggered": "list_targets",
              "expected_behavior": "Lists the workspace's one target, the Support assistant, with when it was last tested and how that run ended."
            },
            {
              "description": "Read the most severe open finding, with the attack and the agent's answer.",
              "prompt": "Show me the most severe open finding on my Support assistant, with the attack and the agent's answer.",
              "tools_triggered": "list_findings, get_finding",
              "expected_behavior": "Names one of the two high-severity open findings on the Support assistant, a cross-session leak or output injection through the session id, and shows its attack transcript: what was sent and how the agent answered."
            },
            {
              "description": "Start an assessment scoped to what a change touched.",
              "prompt": "I changed my Support assistant's system prompt. Run an Enoki assessment on the categories that change could affect.",
              "tools_triggered": "list_categories, trigger_assessment",
              "expected_behavior": "Picks the categories that match a system prompt change, starts an assessment on them, and returns its run id with status running."
            },
            {
              "description": "Follow a running assessment.",
              "prompt": "How far along is the Enoki run I just started?",
              "tools_triggered": "list_runs, get_assessment_status",
              "expected_behavior": "Reports the run's status and the categories completed so far."
            },
            {
              "description": "Check whether a fix worked.",
              "prompt": "I fixed the cross-session leak Enoki found in my Support assistant. Check that it is fixed.",
              "tools_triggered": "list_findings, close_finding, validate_fix",
              "expected_behavior": "Marks the finding fixed, re-runs its attacks against the Support assistant and reports the verdict. The agent has not actually changed, so the verdict is still broken, or validating while the re-run is in progress."
            }
          ],
          "negative": [
            {
              "description": "A general question about a security concept needs no workspace action.",
              "prompt": "What is prompt injection?"
            },
            {
              "description": "Writing a prompt is not a test of an agent.",
              "prompt": "Write me a system prompt for a customer support bot."
            },
            {
              "description": "Scanning a website's ports is not testing an AI agent.",
              "prompt": "Scan example.com for open ports."
            }
          ]
        },
        "demo_recording_url": "https://drive.google.com/file/d/1LUhxcsZ7pm3YF-_HWwEuzDm90Pua8xsc/view?usp=sharing"
      },
      "publication": {
        "release_notes": "First release: register AI agents, run security and safety assessments, read findings with the exact attack, and re-verify fixes."
      },
      "apps": "./.app.json"
    }
  }
}