Instrument Skeptic — Development agent for Claude Code
Adversarially reviews a measurement instrument before its numbers are believed — checks for the known ways a run produces a clean and wrong number.
How to install Instrument Skeptic
Installs to ~/.claude/agents/evolvingagentslabs-lora-kernel-instrument-skeptic.md
mkdir -p ~/.claude/agents && curl -fsSL https://raw.githubusercontent.com/EvolvingAgentsLabs/lora-kernel/HEAD/.claude/agents/instrument-skeptic.md -o ~/.claude/agents/evolvingagentslabs-lora-kernel-instrument-skeptic.md Restart Claude Code, or start a new session, for it to be picked up.
What Instrument Skeptic does
name: instrument-skeptic description: Adversarially reviews a measurement instrument before its numbers are believed — checks for the known ways a run produces a clean and wrong number. Use after building or changing anything that produces a metric, and before any result is written into docs/PLAN.md as established. tools: Read, Bash, Grep, Glob model: sonnet
You distrust the first result. Every failure listed below produced a number that looked like evidence; none was caught by reading
Alternatives in Development
- Get Commit ID — gh api repos/OWNER/REPO/pulls/PR_NUMBER --jq '.head.sha' gh api repos/OWNER/REPO/pulls/PR_NUMBER/comments \ 1.3k ★
- Revision Executor — 교정 실행 전문가. 진단 보고서 4건을 통합하여 에피소드 파일에 교정을 적용 144 ★
- Peer Reviewer Demographics — A simulated peer reviewer specializing in population representativeness, demographic measurement, and populati 134 ★
Full documentation available on GitHub
View Source RepositoryRelated Agents
Corpus Prover
Proves a change on the maintainer's real local corpus — what number moved, by how much, and that nothing else
Name Reviewer
Name Reviewer — the skeptic. Adversarially evaluates proposed names against the brief's audience posture and c
Perf Specialist
Performance optimization agent. Delegate here for speed, bundle size, Web Vitals, profiling. Measurement-first
Skeptic Verifier Fast
Mid-tier variant of skeptic-verifier, dispatched by judgment-dispatch for low-stakes claims and additional med
Skeptic Verifier
Adversarial refuter for load-bearing claims. Recomputes empirical numbers from raw sources and actively tries
Answer Critic
Adversarially attacks one mock-interview answer - tries to prove it wrong, vague, or unsupported. Spawned by /
Related Skills
ShrinkCode
A Claude Skill that compresses bloated codebases into fewer lines without changing behavior: finds duplicate b
Ralph Output Adversarial Review
Adversarially reviews the artifacts produced by a RALPH run (code/tests/UI/config) against repo policy (testin
AI Design Field Guide
46 ways to work with AI design tools — Claude Design, Figma Make, Stitch, Canva, or anything. The thinking app