benchmark

A comparison tool for testing how different AI coding services handle the same programming task. It measures their answers using criteria such as completeness, accuracy, usefulness, and unsupported claims.

In plain words
What is it for?
Use it to compare providers on a code change, a selected file, or a written prompt. It can also run fixed repeatable tasks, test the generated code, and compare results over time.
Why use it?
It provides a repeatable way to compare coding agents instead of judging them from a single informal example.

Skill for Claude CodeCodex

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 skills/greglas75/zuvo/benchmark
Any agent
npx skills add greglas75/zuvo --skill benchmark
Clone the repo
git clone --depth 1 https://github.com/greglas75/zuvo

Made for: Claude Code, Codex.

Per session 76 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 4,699 The whole file, excluding the scripts and references it only reads on demand.
Security scan B 1 finding. 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 $0.00076 $0.04699
Opus 5 $0.00038 $0.02350
Sonnet 5 $0.00015 $0.00940
Haiku 4.5 $0.00008 $0.00470

Measured 2d ago against content hash d79ff2b73164, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade B, and why

benchmark scanned grade B 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 2d 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.

Asks the agent to reveal its instructionsmediumSystem prompt leakage

Directions to print, repeat or translate the system prompt extract configuration the operator did not intend to expose.

| `--dry-run` | Print prompt + providers without dispatching |
skills/benchmark/SKILL.md · 411 lines

How it starts

The opening of the file, as written. The whole thing — 411 lines — stays where its author put it; the contents beside it link to each section on GitHub.

zuvo:benchmark — Multi-Provider Coding Benchmark

Measures how well different AI coding agents handle a task. Dispatches to all available providers in parallel, collects responses, and uses an opposite-model Claude meta-judge to score on four dimensions (completeness, accuracy, actionability, no_hallucinations). Outputs a ranked leaderboard with cost and time breakdown.

Use corpus mode (--mode corpus) to compare runs across time. The corpus uses fixed OrderService + useSearchProducts tasks — the same prompt every time — so runs from different days are directly comparable.

Argument Parsing

Parse $ARGUMENTS for these flags:

Flag Effect
--diff [ref] Benchmark against a git diff (default: HEAD~1)
--files <path> Benchmark on a file or newline-separated list of files as task input
--prompt <text> Use a literal text prompt as the task (alias: --task)
--mode corpus Use fixed corpus tasks (OrderService + useSearchProducts)
--mode default Use user-provided task (default)
--with-tests Run Round 2: have providers write tests for their own Round 1 code
--with-adversarial Run adversarial cross-review on Round 1 code
--with-test-adversarial Run adversarial cross-review on Round 3 tests (requires --with-tests)
--with-static-checks Run tsc + jest on generated code (best-effort, null if tools missing)
--provider <name> Restrict to one or more providers, comma-separated (alias: --providers)
--show-costs Print provider cost table ($/M tokens) and exit
--compare [id1] [id2] Compare two prior runs from zuvo/reports/; default: last two (skill-only, not passed to runner)
--replay-last Re-run benchmark with the same task as the most recent run (skill-only, not passed to runner)
--json Output raw JSON to stdout instead of formatted leaderboard (handled by both runner and skill)
--no-snapshot Suppress task_snapshot storage (use if task contains secrets or PII)
--dry-run Print prompt + providers without dispatching
(remaining text) Treated as --prompt <text> if no other input flag given

Read the full file on GitHub · 411 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. 2d ago First seen · 411 lines · 76 tokens per session scan B d79ff2b73164

Subscribe to this mod's changes

benchmark is a skill published in the GitHub repository greglas75/zuvo (6 stars, last pushed 2d ago), licensed MIT. It adds 76 tokens to every session and 4,699 once invoked, about $0.0004 per session on Opus 5. A static security scan graded it B with 1 finding (asks the agent to reveal its instructions). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.

Related

Other skills, from other repositories

ring:adopting-lib-commons-huma-wrapper

Adopting the lib-commons/v5 shared Huma (OAS 3.1) OpenAPI wrapper + RFC 9457 problem model (commons/net/http/{openapi,problem}) in a Lerian Go service: wire openapi.New/ServeSpec + problem.Install (central >=500 scrub) on BOTH runtime and spec-gen paths, the per-rail problem.MapError flex seam, and rename-only spec…

LerianStudio/ring · 142 tokens

ring:applying-composition-patterns

React composition patterns that scale. Avoid boolean prop proliferation by using compound components, lifting state, and composing internals. Use when refactoring components with boolean prop proliferation, building flexible component libraries, or during architecture review. Skip for simple components with 1-2 props…

LerianStudio/ring · 68 tokens

ring:searching-code

Forensic code search and analysis with optional Chain of Draft (CoD) ultra-concise mode. Five-phase methodology (clarification, planning, execution, analysis, synthesis) with severity assessment. Use for targeted investigation of specific patterns, bugs, or vulnerabilities. Skip for broad architecture mapping (use…

LerianStudio/ring · 74 tokens

ring:exploring-codebases

Exploring a codebase across phases: scopes the target, detects architecture, components, and layers, deep-dives each discovered perspective, then synthesizes findings into actionable guidance with file:line evidence. Use to understand how a feature or system works before planning changes, or to orient on an unfamiliar…

LerianStudio/ring · 91 tokens

ring:opening-pull-requests

Open a GitHub Pull Request with automatic base branch detection, scope allowlist enforcement, PR template filling, and post-create base verification. Replaces ring:generating-pr-descriptions. Use after pushing a branch when ready to open a PR. Skip if the branch is not yet pushed or there are uncommitted changes …

LerianStudio/ring · 83 tokens

ring:cleaning-comments

Cleaning redundant and obvious comments following clean code principles while preserving meaningful documentation. Supports git scope filtering (staged, unstaged, branch, commit-range). Use when code has excessive comments, during code review, or post-refactor cleanup. Skip when reviewing documentation files or…

LerianStudio/ring · 64 tokens