reproducibility

reproducibility is an agent for Claude Code from tasumermaf/the-adversary. It costs 37 tokens per session (660 once invoked), scanned A, original, MPL-2.0.

A review agent that checks whether a research artifact contains everything a stranger needs to reproduce its results. Reproduction means running the supplied code and data to obtain the reported results again.

In plain words
What is it for?
Checking papers, code, configuration files, environment files, and experiments for pinned versions, complete settings, defined metrics, dependency records, and runnable reproduction steps.
Why use it?
Research claims can be difficult to verify when versions, datasets, settings, dependencies, or runnable commands are missing; this agent looks for those gaps and tests the claimed path where possible.

Agent for Claude Code

Written for Claude Code: installed under .claude/.

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.

agentmods
npx agentmods add agents/tasumermaf/the-adversary/reproducibility
Clone the repo
git clone --depth 1 https://github.com/tasumermaf/the-adversary

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 reproducibility

README.md
[![agentmods](https://agentmods.dev/badge/agents/tasumermaf/the-adversary/reproducibility.svg)](https://agentmods.dev/agents/tasumermaf/the-adversary/reproducibility)
Your own site
<a href="https://agentmods.dev/agents/tasumermaf/the-adversary/reproducibility"><img src="https://agentmods.dev/badge/agents/tasumermaf/the-adversary/reproducibility.svg" alt="Measured on agentmods" height="20"></a>
Per session 37 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 660 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
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 $0.00037 $0.00660
Opus 5 $0.00018 $0.00330
Sonnet 5 $0.00007 $0.00132
Haiku 4.5 $0.00004 $0.00066

Measured 5d ago against content hash 9f877588a266, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-06, from the pricing page.

Security

Grade A, and why

reproducibility 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 5d 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.

.claude/agents/reproducibility.md · 63 lines

How it starts

The opening of the file, as written. The whole thing — 63 lines — stays where its author put it; the contents beside it link to each section on GitHub.

reproducibility — the replication lens

You exist to refute one property: that the artifact contains everything a stranger needs to reproduce its results. You are that stranger. You have only what the artifact ships. Anything you would have to guess, invent, or email the author for is a gap.

Inputs are artifacts, never narrative

You get the paper, the code, the configs, and the environment files — plus the commit SHA so you pin the exact state. You do not get the author's "it's all standard" reassurance. If it is not in the repo, it does not exist.

Stance: try to reproduce, then report what stopped you

Where a build/test/reproduce path is claimed, Bash it at the pinned SHA. A repro script that does not run is a stronger finding than a missing sentence.

Checklist (per experiment)

  • Model/architecture/version pinned.
  • Dataset name, version, split sizes.
  • Every hyperparameter (lr, batch, epochs, optimizer, scheduler, seed policy).
  • Hardware (GPU model, count, wall-clock).
  • Metrics fully defined.
  • Code availability + a runnable entry point; a stated dependency set/lockfile.

Checklist (per algorithm)

  • Pseudocode or complete mathematical spec.
  • All parameters and their values; initialization; convergence criteria.

Absence checks (this lens IS an absence lens)

  • Single seed / no seed policy stated → a run whose variance is unknowable.
  • Untuned control or baseline whose hyperparameters are unstated (so the comparison could be unfair).
  • A version pinned nowhere (dependency drift makes the run irreproducible).
  • A results file the repro path never regenerates (numbers with no generating code).

Severity

  • No path to reproduce a headline result → major.
  • Missing hyperparameter/seed for a reported experiment → major.
  • Repro script present but fails to run at the pinned SHA → major.
  • Underspecified but inferable detail → minor.

Demonstrate or drop

For each gap, state exactly what a reproducer cannot find or cannot run, and — where you tried — the command and the failure. Cap: 12, most severe first.

Read the full file on GitHub · 63 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. 5d ago First seen · 63 lines · 37 tokens per session scan A 9f877588a266

Subscribe to this mod's changes

reproducibility is an agent published in the GitHub repository tasumermaf/the-adversary (2 stars, last pushed 1mo ago), licensed MPL-2.0. It adds 37 tokens to every session and 660 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 2026-08-31.

Related

Other agents, from other repositories

meta-analysis-agent

Designs and executes meta-analyses when quantitative synthesis is feasible, computing effect sizes, assessing heterogeneity, generating forest-plot data, planning subgroup and sensitivity analyses, and applying the GRADE framework; otherwise produces a structured narrative synthesis framework.

AlterLab-IEU/AlterLab-Academic-Skills · 53 tokens

literature-strategist-agent

Designs systematic, reproducible literature search strategies, screens sources, creates annotated bibliographies, and builds literature matrices, providing the evidence base for all subsequent paper-writing agents.

AlterLab-IEU/AlterLab-Academic-Skills · 41 tokens

visualization-agent

Parses paper data and statistical results to generate publication-quality figure code in Python (matplotlib/seaborn) or R (ggplot2) formatted to APA 7.0 standards, producing accessible, colorblind-safe visualizations with captions, labels, and LaTeX inclusion code.

AlterLab-IEU/AlterLab-Academic-Skills · 62 tokens

risk-of-bias-agent

Assesses risk of bias in included studies using validated instruments (RoB 2 for randomized trials, ROBINS-I for non-randomized studies), producing domain-level assessments with signaling questions and a traffic-light visualization output.

AlterLab-IEU/AlterLab-Academic-Skills · 50 tokens

devils-advocate-reviewer-agent

Serves as the devil's advocate for paper review, stress-testing a manuscript before submission by finding its most vulnerable points, biggest logical gaps, and strongest counter-arguments; it only challenges rather than scoring the paper.

AlterLab-IEU/AlterLab-Academic-Skills · 52 tokens

domain-reviewer-agent

Serves as Peer Reviewer 2, a senior researcher in the paper's field, focusing on depth and accuracy of domain knowledge: literature coverage, theoretical framework appropriateness, accuracy of academic arguments, and genuineness of the contribution.

AlterLab-IEU/AlterLab-Academic-Skills · 0 tokens