Question 1

Why should I trust a hook to advise me on my own questions?

Accepted Answer

Advising is the default and you still answer; the panel shows a probability per option and a confidence figure, and anything under the 0.75 floor is handed back to you. The README argues for that higher floor on the decision router, since a confidently wrong answer takes a question that was yours to answer.

Question 2

What happens if the API is down or the key is missing?

Accepted Answer

Every path fails open. A missing key, an HTTP error, a timeout, a hook that throws, a body that will not parse and an incomplete skill list all fall through to next(e), and the turn carries on untouched. Each call also races a 4-second sleep, because a hook cannot catch the engine's 10-second dispatch budget.

Question 3

How much does this cost per turn?

Accepted Answer

Jev bills input tokens only, at $0.042 per million, with no decoding charge. The README prices one decision at 374 input tokens and one skill-router prompt at 6,772. The decision router fires only when the agent stops to ask you something, so it costs almost nothing in practice.

Question 4

Why is TYPESAFE_AUTO_ANSWER=1 flagged as a warning?

Accepted Answer

Answering a tool call from a hook requires deny, and the engine defines deny as the model receiving the text as an error result, so a decision that worked renders in red and the agent may argue with it. Returning a result would render normally, but core validates it against the tool's output schema, which the generated types do not declare.

Question 5

How do I install typesafe-mod?

Accepted Answer

Add the marketplace and install the plugin: claude plugin marketplace add BeLazy167/typesafe-mod, then claude plugin install typesafe-mod@typesafe-mod. To run from a checkout instead, use claude --plugin-dir /path/to/typesafe-mod. It also needs early-access function hooks via CLAUDE_CODE_ENABLE_FUNCTION_HOOKS=1 and a TypeSafe API key, or it loads and does nothing.

Question 6

Why is the skill router off by default?

Accepted Answer

Latency, not cost. It runs on every prompt and adds 166 to 420 ms to each one; at 100 prompts a day it would cost about $0.85 a month, but the added delay is the reason it ships disabled. Turn it on with TYPESAFE_SKILL_ROUTER=1.

Question 7

How does it handle a dialog with several questions?

Accepted Answer

All questions ride one request and each step is judged separately, so it can be confident about one and hand the next back to you. Multi-select steps are skipped because a multi-select answer is not a Choice, and the panel draws rows for the first answered step only, capped at four.

TypeSafe Mod

About this project

Highlights

Questions about this entry

More Jev resources like this

From the Jev community