Jev

Add jev-latest (TypeSafe Jev) as a decision helper/tool for my coding agent.
Keep my coding agent on its current chat model. Jev is not a primary coding/chat model and does not generate chat completions.

Optional official skill setup (ask before installing; choose one route):
For Claude Code:
claude plugin marketplace add typesafe-ai/skills
claude plugin install typesafe@typesafe-ai
Then invoke /typesafe:typesafe-ai.
For other supported agents:
npx skills add typesafe-ai/skills --skill typesafe-ai
Then ask the agent to use the TypeSafe skill.
Skill reference: https://docs.typesafe.ai/agent-skill

The vendor skill/examples assume a direct TypeSafe key and origin. Override both for this integration:
- POST only to https://api.staging.experientiallabs.ai/v1/systemone.
- Read the Bearer key only from EXPERIENTIAL_API_KEY. If missing, stop and ask me to set it locally; never ask me to paste or log it.
- Do not request a TypeSafe API key or call the provider's endpoint. If the skill cannot use the gateway, write a small native HTTP helper instead.
- Send model "jev-latest" exactly, preserving any selected :free spelling.

Implement a native JSON request with model, state (text, object, or array), and questions (a map keyed by my question IDs). Each question needs type and instructions:
- choice: criteria maps options such as accept/review to their descriptions.
- noul: judge a proposition; optional criteria describe true and false.
- score: criteria is an ordered list of level descriptions.
Use a concise diff and test summary as state, not the entire repository. Ask independent questions; an answer is not fed into another question in the same request.
Approximately 32,000 input tokens shared by state and all question definitions; not a total context window or an output limit.
Gateway limits: at most 32 questions per request, 64 options per choice, and 2-10 levels per score. These are platform admission bounds, not provider limits.
Native request reference: https://docs.typesafe.ai/introduction/quickstart
Input-budget source: https://docs.typesafe.ai/primitives

Read answers by question ID (for example answers.review), and show the decision and reported usage to me. Probabilities and confidence are not guarantees of correctness. Never execute a returned choice, merge a change, or bypass review automatically.
Use a bounded timeout, surface HTTP and transport errors, and do not add automatic retries or Idempotency-Key: an unknown outcome may already be charged. This endpoint does not stream and has no Chat Completions, Responses, or Messages facade.
Prepare the helper and a mocked test first. Ask before making a potentially charged live call; do not claim it worked without a real response. Tell me which files you changed.