jev-agent-toolkitlisted

Build with Jev, TypeSafe's System One model, as a bounded judgment layer alongside general-purpose agents. Use when calling the TypeSafe API or SDKs, when designing Choice, Score or Noul questions, when classifying, routing, ranking or filtering with confidence thresholds, or when the user mentions Jev, TypeSafe, System One, or bounded semantic judgment. Also covers capability-driven orchestration of external tools (browser research, MCP servers, Blender, Unreal Engine) and multi-agent delegation with sequential fallback.
reiswaffel78/jev-agent-toolkit · ★ 2 · AI & Automation · score 72

Install: claude install-skill reiswaffel78/jev-agent-toolkit

# Jev Agent Toolkit Jev is a **System One** model from TypeSafe. Give it a `state` and typed `questions`; it returns typed answers with calibrated probability distributions. **Jev does not generate text or code.** That stays with the general-purpose agent running this skill. Jev supplies bounded judgment; deterministic code supplies exact operations and control flow; external tools perform actions. ``` JEV · bounded judgment classification / score / probability | this skill (portable core) | lead agent / orchestrator | +-----------------+-----------------+ CODE BROWSER TOOLS | +--------------+--------------+ BLENDER UNREAL FUTURE TOOLS ``` Do not force work through Jev to keep it central. Keeping deterministic things deterministic *is* the architecture. ## Division of labour | Worker | Owns | |---|---| | Deterministic code | Arithmetic, counting, dates, parsing, file I/O, builds, thresholds, policy, control flow | | Jev | Bounded semantic judgment with a probability attached | | General-purpose LLM | Open-ended reasoning, writing prose and code, generating candidates | | External tools | Actions in the world | ## The decision gate Bef