What is bench-creator?

A tool for collecting reusable examples of real coding tasks in a private benchmark. A benchmark is a repeatable test used to compare how well coding agents handle difficult work.

Use it before and after substantive repository changes to capture benchmark cases and run them in isolated, restored workspaces.

It preserves the starting code, task difficulty, and evaluation setup so future agent changes can be tested consistently and privacy rules can be checked.

bench-creator — Skill for Claude Code and Codex by EvanLuo42

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 EvanLuo42/bench-creator --skill bench-creator

Clone the repo

git clone --depth 1 https://github.com/EvanLuo42/bench-creator

Made for: Claude Code, Codex.

Or install bench-creator, the plugin that ships this one along with the rest of its 1 skill.

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.

README.md

[![agentmods](https://agentmods.dev/badge/skills/evanluo42/bench-creator/bench-creator/github.svg)](https://agentmods.dev/skills/evanluo42/bench-creator/bench-creator)

Your own site

<a href="https://agentmods.dev/skills/evanluo42/bench-creator/bench-creator"><img src="https://agentmods.dev/badge/skills/evanluo42/bench-creator/bench-creator/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.

Your own site · 80×15

<a href="https://agentmods.dev/skills/evanluo42/bench-creator/bench-creator"><img src="https://agentmods.dev/badge/skills/evanluo42/bench-creator/bench-creator.svg" alt="Reviewed on agentmods" width="80" height="20"></a>

Per session 116 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.

When invoked 900 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.

Model	Per session	Once invoked
Fable 5.1	$0.00116	$0.00900
Opus 5.5	$0.00046	$0.00360
Sonnet 5.5	$0.00023	$0.00180
Haiku 4.5	$0.00012	$0.00090

Measured 2mo ago against content hash 10bbaa0ef0ec, method: parsed. Prices are Anthropic first-party input rates as of 2026-10-07, from the pricing page.

Security

Grade A, and why

bench-creator 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 2mo ago.

The scan reads SKILL.md. This mod also ships 5 executable files (scripts/benchctl.ps1, scripts/benchctl.py, scripts/benchctl.sh, …), listed below but not scanned — reading those needs a real analyzer, not pattern matching.

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.

skills/bench-creator/SKILL.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.

Capture Bench Cases

Maintain one global ai-work-bench/v1 hub. Treat repositories as sources, not Bench containers.

Resolve the command and hub

Resolve the absolute directory containing this SKILL.md; never assume the current working directory is the Skill directory.

Windows: run powershell -NoProfile -ExecutionPolicy Bypass -File <skill-root>/scripts/benchctl.ps1 ...
POSIX: run sh <skill-root>/scripts/benchctl.sh ...

Use <benchctl> below as shorthand for that absolute launcher command. Set AI_WORK_BENCH_PYTHON only when an explicit Python interpreter override is needed.

Resolve the Bench from AI_WORK_BENCH_HOME, then $CODEX_HOME/benches/daily-work, then ~/.codex/benches/daily-work. Do not create a per-repository .bench. If bench.json is missing, initialize only when the user explicitly requests setup:

<benchctl> init --bench <global-path> --name "Daily Work Bench" --mode suggest --workspace-mode snapshot

Before repository work

Ask the CLI for the policy decision before the first edit or generated artifact:

<benchctl> policy decide --bench <global-path> --phase before

skip: continue without Bench work.
checkpoint: read references/checkpointing.md, then run checkpoint start with the returned trigger.
For a user-requested capture, add --explicit to the decision and use trigger explicit.

Keep the returned checkpoint and task IDs. If checkpointing fails, continue the parent task without capture and never reconstruct input from the solved workspace unless a verified historical parent exists.

After repository work

Collect only signals that actually occurred, then ask for the final decision:

<benchctl> policy decide --bench <global-path> --phase after --signal <signal> ...

discard: run checkpoint discard.
suggest: ask for confirmation; capture with --confirmed only after approval.
capture: read references/case-authoring.md, create the case, and attach the checkpoint.

Read the full file on GitHub · 63 lines

Files

What ships with it

13 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.

2mo ago First seen · 63 lines · 116 tokens per session scan A 10bbaa0ef0ec

Subscribe to this mod's changes

bench-creator is a skill published in the GitHub repository EvanLuo42/bench-creator (5 stars, last pushed 3mo ago), licensed MIT. It adds 116 tokens to every session and 900 once invoked, about $0.0005 per session on Opus 5.5. 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.

Building agents Claude Code Testing agent codex skills

bench-creator next to 6 skills from other repositories
Mod	Type	Stars	Last push	Licence	Scan	Tokens	Works with
bench-creatorEvanLuo42/bench-creator	Skill	5	3mo ago	MIT	A	116	Claude Code, Codex
make_planagentscope-ai/QwenPaw	Skill	37k	2d ago	Apache-2.0	A	51	Claude Code, Codex
motion-showreelConardLi/garden-skills	Skill	15k	14d ago	MIT	A	107	Claude Code, Codex
gpt-image-2ConardLi/garden-skills	Skill	15k	14d ago	MIT	A	177	Claude Code, Codex
newclacky-ai/openclacky	Skill	1.2k	2d ago	MIT	A	10	Claude Code
oss-uploadclacky-ai/openclacky	Skill	1.2k	2d ago	MIT	A	40	Claude Code, Codex
windbg-user-heap-corruption-investigationmicrosoft/win-dev-skills	Skill	501	6d ago	MIT	A	46	Claude Code