Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIDeepSeekDeepSeekMuseMuse+22 more

    JEV Calibrated Decisions for AI Agents

    1

    Ground bounded decisions in supplied facts using TypeSafe JEV MCP probabilities. Portable guidance for Claude Code, Codex, Cursor, Copilot and other MCP-capable agents.

    $7

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    JEV Calibrated Decisions for AI Agents

    JEV Calibrated Decisions for AI Agents

    Example session with this skill installed

    Use JEV to evaluate whether this change is ready to proceed without review. State: change_is_reversible=false; review_complete=false. Show the actual probability and explain its scope.

    • Read your context and instructions
    • Compiled the jev calibrated decisions

    A live evaluation of the packaged server returned:

    {"probability": 0.06}
    

    Proposition evaluated: "The change is ready to proceed without review."

    Supplied state

    {"change_is_reversible": false, "review_complete": false}
    

    Interpretation: JEV assigned a 6% probability to the proposition from these two supplied facts. This is a model estimate, not proof that the change is safe or unsafe. No authorization threshold was supplied, so the skill does not invent one or declare the change approved. No external change was executed by JEV.

    Next step: obtain the missing review evidence and evaluate the decision again if needed, while retaining the host agent’s existing permissions. This example was executed against TypeSafe using the local MCP server; results may vary between model versions and requests. Automatic Codex tool gating requires the separate trusted hook configuration.

    Connects securely to your tools. The creator never sees your data.

    About this skill

    Use TypeSafe JEV as a typed decision assistant for eligibility checks, evidence-based verification, classification, and candidate-tool judgments.

    The skill guides agents to gather relevant state, formulate a precise yes/no proposition, invoke the connected jev_noul MCP tool, and show the returned probability without claiming certainty.

    Includes portable instructions and an exact MCP input example. Requires a separately installed JEV MCP server and your own TypeSafe API key. TypeSafe API usage is billed separately; this purchase includes no API credits.

    Works with agents that can load SKILL.md instructions and invoke the connected JEV MCP tool. The optional automatic PreToolUse guard is specific to supported Codex installations and must be configured separately. No promise of universal interception, faster execution, or reduced token costs.

    Source and setup: https://github.com/Idk507/jev-codex-plugin

    Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for LLM Ops.

    Requires separate JEV MCP server setup and TypeSafe API credits. It does not install Python or the JEV plugin. Probabilities are estimates, not guarantees.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before buying a skill.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    Fresh listing

    Recently published to Agensi

    Money-back guarantee

    Doesn't do what its page promises? You get your money back.

    Trust & safety

    Security scan

    Security scan passed 13 days ago

    Accounts
    TypeSafe API key
    Software
    JEV MCPPython 3.11+

    Install and connect the JEV MCP separately; Python runs it locally. TypeSafe API charges are separate.

    Permissions

    Network Access
    Write Files
    Environment Variables

    Allowed Hosts

    typesafe.ai

    The connected JEV MCP server sends the supplied decision state and proposition to TypeSafe. API keys stay in the separately configured local server; the skill does not request or transmit credential values. No external actions are executed by the skill.

    Listed13 days ago

    What's inside

    Frequently Asked Questions

    JEV returns a typed probability for a specific proposition over supplied facts. It can still be wrong; the estimate does not eliminate hallucination or replace independent evidence.

    The instructions are portable to Claude Code, Codex, Cursor, Copilot and other hosts that can load skills and invoke the connected JEV MCP tool. Each host needs separate MCP setup; the package has not been tested independently in every host.

    You will need an active TypeSafe API account and the JEV MCP server installed in your environment. Once these are configured, you can call the `jev_noul` tool by passing it a JSON state object and the specific proposition you need to evaluate.

    Supported local Codex tool calls can be evaluated automatically after configuring and trusting the separate PreToolUse MCP hook. The skill alone does not install hooks, intercept arbitrary actions, or guarantee cross-agent automatic gating.

    JEV Decisions provides bounded estimates based on the evidence supplied, but it is not a formal mathematical proof. Additionally, it does not support normalized ranking across multiple separate calls; each proposition is evaluated independently.

    New in LLM Ops