flowreg-benchmark-reproduce

flowreg-benchmark-reproduce is a skill for Claude Code from FlowRegSuite/flowreg-agent-skills. It costs 108 tokens per session (2,969 once invoked), scanned A, original, MIT.

A workflow for reproducing or auditing published motion-correction benchmark results using fixed source versions and documented settings. A benchmark is a controlled test used to compare methods or measure performance.

In plain words
What is it for?
Use it to regenerate benchmark numbers, verify published figures, check comparisons with NoRMCorre, or answer reproducibility questions.
Why use it?
It helps determine whether a reported result was reproduced exactly, approximated, or affected by different parameters, code revisions, coordinate conventions, or comparison settings.

Skill for Claude Code

Written for Claude Code: shipped in a Claude Code plugin.

Part of the flowreg-agent-skills plugin — 7 skills shipped together

Good fit Use it to regenerate benchmark numbers, verify published figures, check comparisons with NoRMCorre, or answer reproducibility questions.

Compare 6 skills from other repositories ↓
Install with agentmods
npx agentmods add skills/flowregsuite/flowreg-agent-skills/flowreg-benchmark-reproduce
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.

Any agent
npx skills add FlowRegSuite/flowreg-agent-skills --skill flowreg-benchmark-reproduce
Clone the repo
git clone --depth 1 https://github.com/FlowRegSuite/flowreg-agent-skills

Made for: Claude Code.

Or install flowreg-agent-skills, the plugin that ships this one along with the rest of its 7 skills.

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 flowreg-benchmark-reproduce

README.md
[![agentmods](https://agentmods.dev/badge/skills/flowregsuite/flowreg-agent-skills/flowreg-benchmark-reproduce/github.svg)](https://agentmods.dev/skills/flowregsuite/flowreg-agent-skills/flowreg-benchmark-reproduce)
Your own site
<a href="https://agentmods.dev/skills/flowregsuite/flowreg-agent-skills/flowreg-benchmark-reproduce"><img src="https://agentmods.dev/badge/skills/flowregsuite/flowreg-agent-skills/flowreg-benchmark-reproduce/github.svg" alt="Measured on agentmods" height="20"></a>

Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.

agentmods 80×15 button for flowreg-benchmark-reproduce

Your own site · 80×15
<a href="https://agentmods.dev/skills/flowregsuite/flowreg-agent-skills/flowreg-benchmark-reproduce"><img src="https://agentmods.dev/badge/skills/flowregsuite/flowreg-agent-skills/flowreg-benchmark-reproduce.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 108 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 2,969 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 1 finding. 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 $0.00108 $0.02969
Opus 5 $0.00054 $0.01484
Sonnet 5 $0.00022 $0.00594
Haiku 4.5 $0.00011 $0.00297

Measured 11d ago against content hash 6de480f63b2d, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-11, from the pricing page.

Security

Grade A, and why

flowreg-benchmark-reproduce scanned grade A with 1 finding 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 11d 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.

Makes network callslowCapability

Not a fault in itself. Listed so you know the mod talks to something, and to what.

- It reports neither the 95th-percentile error, the `curl` column, nor the sign guard.
skills/flowreg-benchmark-reproduce/SKILL.md · 232 lines

How it starts

The opening of the file, as written. The whole thing — 232 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Benchmark reproduction and audit

1. Purpose

Select the right benchmark parameter profile, resolve every value to its source revision, and run the benchmark using pyflowreg's own implementation — then state honestly whether the result is an exact reproduction or an approximation.

The failure this skill exists to prevent is a plausible benchmark table whose parameters came from a paper's prose, a stale comment, or a different revision than the one claimed.

This skill selects parameters and records provenance. It does not ship its own benchmark implementation: the displacement field, the noise conditions and the end-point-error metric belong to pyflowreg and its benchmark repository.

2. Use this skill when

  • A benchmark number must be regenerated or independently checked.
  • Someone asks which parameters produced a published figure.
  • A comparison against NoRMCorre needs its exact rigid or non-rigid settings.
  • A reviewer asks whether a result is reproducible, and the answer must be defensible.

3. Do not use this skill when

  • You are correcting real experimental data → flowreg-plan, then the execution skill.
  • You are assessing a finished correction run → flowreg-qc-and-audit.
  • You want new parameters tuned for your own data. Benchmark presets are pinned historical artifacts, not recommendations; flowreg-plan owns parameter selection.

4. Required inputs

  • The condition to reproduce: synthetic (clean, noisy35db, noisy30db) with one or two channels, or a MATLAB reproduction script at 6.2 Hz or 30.9 Hz.
  • A writable run directory outside any source checkout.
  • For the synthetic benchmark: an installed pyflowreg (section 6, step 4).
  • For MATLAB profiles: a MATLAB installation. This skill never executes MATLAB; it extracts parameters statically and reports them.

5. Preflight checks

  1. Identify the condition before opening the catalog. Choosing a profile first and then deciding what it reproduces is how classes get mixed.
  2. Confirm the runtime: flowreg-agent-skills doctor --json reports whether pyflowreg is importable and at which version.
  3. Verify the dataset hash against the catalog. A mismatch downgrades the run to approximate at best, and usually means a different file.
  4. Confirm the coordinate and component conventions (section 7) before computing any metric.
  5. Decide up front what would make this run approximate. Discovering it afterwards is how approximations get published as exact.

Read the full file on GitHub · 232 lines

Files

What ships with it

6 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.

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. 11d ago First seen · 232 lines · 108 tokens per session scan A 6de480f63b2d

Subscribe to this mod's changes

flowreg-benchmark-reproduce is a skill published in the GitHub repository FlowRegSuite/flowreg-agent-skills (2 stars, last pushed 1mo ago), licensed MIT. It adds 108 tokens to every session and 2,969 once invoked, about $0.0005 per session on Opus 5. A static security scan graded it A with 1 finding (makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.