Klein Referee — Research agent for Claude Code
REFEREE worker for Klein Auto Research — Gate 3. Reviews a synthesized study in a FRESH context on a different model than the experimenter, runs the mechanical verifiers (klein verify --numbers --evid.
How to install Klein Referee
Installs to ~/.claude/agents/xiang-shan-klein-auto-research-klein-referee.md
mkdir -p ~/.claude/agents && curl -fsSL https://raw.githubusercontent.com/Xiang-Shan/klein-auto-research/HEAD/.claude/agents/klein-referee.md -o ~/.claude/agents/xiang-shan-klein-auto-research-klein-referee.md Restart Claude Code, or start a new session, for it to be picked up.
What Klein Referee does
name: klein-referee description: REFEREE worker for Klein Auto Research — Gate 3. Reviews a synthesized study in a FRESH context on a different model than the experimenter, runs the mechanical verifiers (klein verify --numbers --evidence-use, klein claims verify, figure re-render), applies the fixed ten-check rubric, and writes referee_report.md with a machine-read Verdict line. Invoke to "referee the study", "review the findings", "run the referee gate", or before any "finalize" on a schema
Alternatives in Research
- Strategist Critic — Empirical strategy critic and gatekeeper 250 ★
- Deep Research Writer — Writes the single report.md of a nav-deep-research run from the run's source notes, following the readable rep 232 ★
- Study — Builds one optional provisional HTML study inside an explicit host-granted directory 190 ★
Full documentation available on GitHub
View Source RepositoryRelated Agents
Klein Experimenter
EXPERIMENT-loop worker for Klein Auto Research — runs disciplined one-candidate v2 transactions with immutable
Klein Sweeper
SWEEP worker for Klein Auto Research — runs the one sanctioned mechanical parameter search, retaining every tr
Doc Drift Auditor
Verify documentation accuracy against implementation using git forensics and code analysis with file paths, li
Kb Corpus Curator
Register, ingest and verify a source in this knowledge base through the kb- mise tasks, and close the loop wit
Permit Websearch
Runs a single WEBSEARCH follow-up proposed during permit extraction and writes a synthesized Answer/Findings/G
Ba Critic
Separate-context business-analysis reviewer dispatched by the ba-orchestrator skill. Reviews one method artifa
Related Skills
Learn Skill
Agent skill that distills AI study chats into a causal-chain learning wiki — mechanical checks, a regression e
Ideonomy Skill
A Claude Code skill for expanding ideas via Patrick Gunkel's ideonomy, synthesized from Grace Kind's essays. P
Santoryu Skills
Three blades, one mind. Opus (Claude Code) plans and reviews; two fast "blades" — a repo-aware Cursor agent an