experiment-reviewer

experiment-reviewer is an agent for Claude Code from revfactory/harness-100. It costs 36 tokens per session (814 once invoked), scanned A, original, Apache-2.0.

A quality reviewer for machine-learning experiments. It checks whether data, model training, evaluation, and conclusions are consistent, scientifically sound, and reproducible.

In plain words
What is it for?
It is for reviewing experiment design, checking reproducibility, validating evaluation results, and suggesting corrections.
Why use it?
It helps catch problems such as data leakage, unfair comparisons, overfitting, or conclusions that the results do not support.

Agent for Claude Code

Written for Claude Code: installed under .claude/.

Good fit It is for reviewing experiment design, checking reproducibility, validating evaluation results, and…

Compare 6 agents from other repositories ↓
Install with agentmods
npx agentmods add agents/revfactory/harness-100/experiment-reviewer
About the project

Harness 100 is a collection of ready-to-use Claude Code agent teams, with specialist agents, orchestrator skills, and domain-specific extensions across many types of work. It is for assembling coordinated agent workflows for software, content, business, education, and other tasks. The catalogue entries are examples of the agents in this collection.

revfactory/harness-100 · 1,259 stars · on GitHub

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

Clone the repo
git clone --depth 1 https://github.com/revfactory/harness-100

Made for: Claude Code.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for experiment-reviewer

README.md
[![agentmods](https://agentmods.dev/badge/agents/revfactory/harness-100/experiment-reviewer.svg)](https://agentmods.dev/agents/revfactory/harness-100/experiment-reviewer)
Your own site
<a href="https://agentmods.dev/agents/revfactory/harness-100/experiment-reviewer"><img src="https://agentmods.dev/badge/agents/revfactory/harness-100/experiment-reviewer.svg" alt="Measured on agentmods" height="20"></a>
Per session 36 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 814 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5.1 $0.00036 $0.00814
Opus 5 $0.00018 $0.00407
Sonnet 5 $0.00007 $0.00163
Haiku 4.5 $0.00004 $0.00081

Measured 3d ago against content hash c4acd365a653, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-06, from the pricing page.

Security

Grade A, and why

experiment-reviewer scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 3d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

en/31-ml-experiment/.claude/agents/experiment-reviewer.md · 94 lines

How it starts

The opening of the file, as written. The whole thing — 94 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Experiment Reviewer — Experiment Reviewer

You are an ML experiment quality verification specialist. You verify the scientific rigor, reproducibility, and validity of conclusions.

Core Responsibilities

  1. Data Leakage Verification: Check whether test data information leaked during preprocessing/feature engineering
  2. Experiment Design Verification: Verify that comparative experiments are fair and statistically significant
  3. Reproducibility Verification: Check that code, data, and environment are recorded for reproducibility
  4. Overfitting Verification: Check whether the performance gap between training and validation is reasonable
  5. Conclusion Validity: Verify that conclusions drawn from evaluation results are supported by data

Working Principles

  • Cross-compare all outputs. Verify consistency across data → model → training → evaluation
  • Evaluate from a paper reviewer's perspective: "Can these experimental results be trusted?"
  • When problems are found, provide specific correction suggestions alongside
  • Classify severity into 3 levels: 🔴 Must fix / 🟡 Recommended fix / 🟢 For reference

Verification Checklist

Data Verification

  • No data leakage (fit on train, transform only on test)
  • Data splitting is appropriate (time-series: chronological, imbalanced: stratified)
  • Preprocessing pipeline is reproducible

Model Verification

  • Compared with baseline model
  • Model complexity is appropriate for data scale
  • Hyperparameter search is systematic

Training Verification

  • Random seeds are fixed
  • No anomalies in training curves (divergence, early overfitting)
  • Checkpoint strategy is appropriate

Evaluation Verification

  • Evaluation metrics are appropriate for the problem
  • Statistical significance is confirmed
  • Error analysis has been performed
  • Bias verification has been performed (when applicable)

Output Format

Save as _workspace/05_review_report.md:

Read the full file on GitHub · 94 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 3d ago First seen · 94 lines · 36 tokens per session scan A c4acd365a653

Subscribe to this mod's changes

experiment-reviewer is an agent published in the GitHub repository revfactory/harness-100 (1,259 stars, last pushed 5mo ago), licensed Apache-2.0. It adds 36 tokens to every session and 814 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-03.

Related

Other agents, from other repositories

nn-embedding-expert

Embedding trained neural networks and tree ensembles as MINLP constraints via discopt.nn - OMLT-style full-space and reduced-space formulations, ReLU big-M, interval bound propagation, ONNX reader. Use when a trained ML surrogate must live inside an optimization problem.

jkitchin/discopt · 59 tokens

structured-data-worker

Generates structured tabular data (encounters, labs, hospitalizations, medications, PROs) for a single patient from their event list and document summaries. Reads table schemas from YAML files. Writes JSON output to a specified path. Spawned by the generate-synthetic-data skill -- do not invoke directly.

kenlkehl/onc-data-wrangler-plugin · 66 tokens

ml-engineer

Use this agent when working with model architecture, training loops, loss functions, optimizers, hyperparameter tuning, experiment tracking, or model evaluation. For example: building a PyTorch model, writing a training loop with mixed precision, setting up an Optuna hyperparameter sweep, configuring MLflow experiment…

morganmuli/metaskill · 81 tokens

bayesian-network-prediction

Constructs and operates probabilistic graphical models for causal inference, prediction under uncertainty, and dynamic belief updating with verified mathematical foundations and real-world integration.

ChrisRoyse/610ClaudeSubagents · 35 tokens

Demonstrate

Agent for demonstrating VS Code features.

microsoft/vscode · 10 tokens

playwright-test-generator

Use this agent when you need to create automated browser tests using Playwright Examples: Context: User wants to generate a test for the test plan item.

microsoft/playwright · 151 tokens