Claude Referee — Development skill for Claude Code
Evidence over eloquence.
How to install Claude Referee
This entry records only its repository, not the path inside it, so there is no
exact command to give. Open ismaildasci/claude-referee and copy the folder into
~/.claude/skills/, or the file into ~/.claude/agents/.
What Claude Referee does
Evidence over eloquence. Unofficial Claude Code plugin that checks "done" claims and small judgement calls with TypeSafe Jev, and keeps receipts.
Alternatives in Development
- Referee 541 ★
- Fast Jev Compaction — Claude Code plugin that replaces the compaction summary with Jev decisions: every tool call and result is scor 425 ★
- Agenfk Plan — Decompose an EPIC into stories (required), or a large STORY into tasks (agent's judgement) 64 ★
README
An unofficial Claude Code plugin that checks "done" claims and small judgement calls with TypeSafe Jev, and keeps receipts.
Install · How it works · What's measured · What it costs · Manifesto (Türkçe) · Measurements · Türkçe
Coding agents write well. "All tests pass. Done." is one fluent line, and it costs nothing to write. Checking it takes a test run, and if the line is wrong, someone finds out later.
claude-referee is an unofficial plugin for Claude Code that checks lines like that. Claude still makes the big decisions. Small questions that can be checked, such as *is it really done?* or *which of these options fits our rules?*, go to Jev: a model from TypeSafe that answers with a probability, like "0.97 yes", instead of a paragraph. Every check is logged on your machine.
What it does
| When | What happens | Release |
|---|---|---|
| You or Claude run a command | done, decide, judge or verify ask |
Related Skills
System One Models Getting Started
TypeSafe's System One launch claims, each as a runnable experiment: Jev vs Claude vs keyword rules through one
Jev Permission Gate
A Claude Code mod that uses TypeSafe's Jev to decide auto mode tool calls. 2x faster than the built-in classif
Jev Skill
Jev skill for Claude Code and Codex: TypeSafe System One guide, Python CLI, 5 tested recipes, and findings fro
Hearmemory
Shared memory for multi-agent coding: subagents and Codex/Claude Code/Cursor hand-offs share one project memor
One Skill To Rule Them All
The meta-skill that builds and improves all your skills, including itself. Watches your work sessions (autonom
Jev Sales Copilot
Live sales-call copilot on TypeSafe Jev: per-utterance closing probability, signals and next-best-move; syncs
Related Agents
Reviewer Specialist
Acts as a demanding specialist referee (collective intelligence / methods) checking claims against evidence. R
Typesafe Adversary
Red-teams TypeSafe integrations before users do. Probes the nine documented Jev failure modes, tests prompt in
Jev Judge
Reusable Judge/Verifier question-drafter for text-shaped deliverables — pass/fail, meets-bar yes/no, or rubric