bench-reviewer

bench-reviewer is an agent for Claude Code from zivtech/accessibility-skills. It costs 35 tokens per session (797 once invoked), scanned A, original, GPL-3.0.

A read-only reviewer for checking the quality of software evaluation tests, including their test data, scoring rules, difficulty, and misleading cases.

In plain words
What is it for?
Use it to inspect evaluation tests, compare fixtures with rubrics, check scoring accuracy, and identify cases that wrongly mark answers as correct or incorrect.
Why use it?
It helps find unreliable evaluations before they lead to incorrect conclusions about a system's performance.

Agent for Claude Code

Written for Claude Code: installed under .claude/. Also seen: model in frontmatter.

Good fit Use it to inspect evaluation tests, compare fixtures with rubrics, check scoring…

Compare 6 agents from other repositories ↓
Install with agentmods
npx agentmods add agents/zivtech/accessibility-skills/bench-reviewer
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/zivtech/accessibility-skills

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 bench-reviewer

README.md
[![agentmods](https://agentmods.dev/badge/agents/zivtech/accessibility-skills/bench-reviewer.svg)](https://agentmods.dev/agents/zivtech/accessibility-skills/bench-reviewer)
Your own site
<a href="https://agentmods.dev/agents/zivtech/accessibility-skills/bench-reviewer"><img src="https://agentmods.dev/badge/agents/zivtech/accessibility-skills/bench-reviewer.svg" alt="Measured on agentmods" height="20"></a>
Per session 35 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 797 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 unknown 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 $0.00035 $0.00797
Opus 5 $0.00017 $0.00398
Sonnet 5 $0.00007 $0.00159
Haiku 4.5 $0.00003 $0.00080

Measured yesterday against content hash afa70b603903, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-06, from the pricing page.

Security

Grade A, and why

bench-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 yesterday.

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/bench-reviewer.md · 69 lines

The source is not reproduced here

Licensed GPL-3.0

The repository is licensed GPL-3.0, which this catalogue does not treat as permission to reproduce the file. Read it at the source.

Read it on GitHub

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. yesterday Changed · +1 lines afa70b603903
  2. 6d ago First seen · 68 lines · 35 tokens per session scan A 5deb104d7e3a

Subscribe to this mod's changes

bench-reviewer is an agent published in the GitHub repository zivtech/accessibility-skills (6 stars, last pushed 2d ago), licensed GPL-3.0. It adds 35 tokens to every session and 797 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

Accessibility Tool Builder

Expert in building accessibility scanning tools, rule engines, document parsers, report generators, and audit automation. Designs WCAG criterion mapping, severity scoring algorithms, CLI/GUI scanner architecture, and CI/CD integration for accessibility tooling.

Community-Access/accessibility-agents · 49 tokens

Desktop Accessibility Specialist

Desktop application accessibility expert -- platform APIs (UI Automation, MSAA/IAccessible2, NSAccessibility), accessible control patterns, screen reader Name/Role/Value/State, focus management, high contrast, and custom widget accessibility for Windows and macOS desktop applications.

Community-Access/accessibility-agents · 58 tokens

Desktop A11y Testing Coach

Desktop accessibility testing expert -- testing with NVDA, JAWS, Narrator, and VoiceOver screen readers, Accessibility Insights for Windows, automated UIA testing, keyboard-only testing flows, high contrast verification, and creating desktop accessibility test plans.

Community-Access/accessibility-agents · 55 tokens

markdown-scanner

Internal helper for scanning a single markdown file for accessibility issues across all 9 domains. Returns structured findings with severity, line numbers, suggested fixes, and auto-fix classification. Invoked by markdown-a11y-assistant via agent tool - not user-invocable directly.

Community-Access/accessibility-agents · 59 tokens

lighthouse-bridge

Internal helper that bridges Lighthouse CI accessibility audit data with the agent ecosystem. Parses Lighthouse reports, normalizes accessibility findings, tracks score regressions, and deduplicates against local scans.

Community-Access/accessibility-agents · 41 tokens

Accessibility Regression Detector

Detects accessibility regressions by comparing audit results across commits, branches, or time periods. Tracks score trends, identifies newly introduced issues, and validates that fixes from previous audits remain in place.

Community-Access/accessibility-agents · 42 tokens