leaderboard: Skill for Claude Code

.claude/skills/benchmark-discovery/SKILL.md

benchmark-discovery is a skill for Claude Code from steel-dev/leaderboard. It costs 246 tokens per session (3,155 once invoked), scanned A, original, MIT.

A workflow for finding recent arXiv papers that may contain results for the Steel benchmark leaderboard. arXiv is a public website where researchers share papers, and a benchmark is a standard test used to compare systems.

In plain words
What is it for?
It helps run the paper-discovery process, search selected benchmarks or date ranges, remove already listed results, and create a Markdown summary for human review.
Why use it?
It filters irrelevant or duplicate papers and produces review material, so maintainers do not have to search and screen every result manually.

Skill for Claude Code

Written for Claude Code: installed under .claude/.

This is steel-dev/leaderboard's own configuration. It tells Claude Code how to work on leaderboard itself, so it is not a mod to install elsewhere. Copy it as a starting point and replace the rules that are about this project. Everything leaderboard configures →

Not installable: its command points at a path on the author’s own machine, so it runs nowhere else. The line is /Users/nikola/dev/steel/leaderboard.

Reuse

Borrowing it

Nothing to install: this file belongs to steel-dev/leaderboard. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.

Copy the file
curl -O https://raw.githubusercontent.com/steel-dev/leaderboard/main/.claude/skills/benchmark-discovery/SKILL.md
Clone the repo
git clone --depth 1 https://github.com/steel-dev/leaderboard

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 benchmark-discovery

README.md
[![agentmods](https://agentmods.dev/badge/skills/steel-dev/leaderboard/benchmark-discovery.svg)](https://agentmods.dev/skills/steel-dev/leaderboard/benchmark-discovery)
Your own site
<a href="https://agentmods.dev/skills/steel-dev/leaderboard/benchmark-discovery"><img src="https://agentmods.dev/badge/skills/steel-dev/leaderboard/benchmark-discovery.svg" alt="Measured on agentmods" height="20"></a>
Per session 246 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 3,155 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.

ModelPer sessionOnce invoked
Fable 5.1 $0.00246 $0.03155
Opus 5 $0.00123 $0.01577
Sonnet 5 $0.00049 $0.00631
Haiku 4.5 $0.00025 $0.00315

Measured 7d ago against content hash 45d4f7155ac7, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-07, from the pricing page.

Security

Grade A, and why

benchmark-discovery 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 7d 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/skills/benchmark-discovery/SKILL.md · 170 lines

How it starts

The opening of the file, as written. The whole thing — 170 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Benchmark discovery for the Steel leaderboard

Surface candidate benchmark results for maintainer review. The skill drives the deterministic arxiv discovery script in this repo, applies project-specific filtering rules, and renders a standardised markdown summary that a maintainer can scan and selectively add to src/data/<slug>.json.

This skill never modifies leaderboard data on its own. The output is review material; merging is always a human step. The discovery issue exists so candidates accumulate in one inbox rather than getting lost across notebooks and tabs.

Process

1. Run the discovery script

The script lives at src/scripts/discover-results.ts and is exposed as npm run discover. It hits arxiv's public API (no auth, ~1 req per 3 seconds per benchmark), filters candidates by date, dedupes by arxiv URL against existing rows, and writes a consolidated .discovery/run.json.

Default invocation:

npm run discover -- --since-days 10

Adjust to user intent:

  • Specific benchmarks: --slug browsecomp,gaia,webvoyager
  • Catching up after weeks off: --since-days 60
  • Available slugs match those declared in src/lib/benchmark-hub.ts (e.g. webvoyager, browsecomp, webarena, swe-bench-verified, osworld, gaia, clawbench, online-mind2web, tau-bench, agentbench).

2. Read .discovery/run.json

Each benchmark entry contains:

  • slug, benchmarkName
  • existingTopSystems — the top 8 rows currently tracked, used for the "Current top" line in the output
  • candidates[] — papers with arxivId, url, title, authors, publishedAt, summary (full abstract)

The script has already dropped anything whose arxiv ID matches an existing sourceUrl. Anything in candidates is genuinely unmatched at the URL level — but may still be a duplicate by system name (many existing rows source to blog posts/HF, not arxiv), or off-topic entirely. Classification is the skill's job.

3. Classify each candidate by reading its abstract

Read the full file on GitHub · 170 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. 7d ago First seen · 170 lines · 246 tokens per session scan A 45d4f7155ac7

Subscribe to this mod's changes

benchmark-discovery is a skill published in the GitHub repository steel-dev/leaderboard (48 stars, last pushed yesterday), licensed MIT. It adds 246 tokens to every session and 3,155 once invoked, about $0.0012 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 2026-08-30.

Related

Other skills, from other repositories

browser-use

Direct browser control via CDP for web interaction: automation, scraping, testing, screenshots, and site/app work.

browser-use/browser-use · 26 tokens

drug-discovery

Drug discovery: ChEMBL search, drug-likeness, interactions.

NousResearch/hermes-agent · 19 tokens

jupyter-notebook

Iterative Python via live Jupyter kernel (hamelnb).

NousResearch/hermes-agent · 18 tokens

open-source

Documentation reference for writing Python code using the browser-use open-source library. Use this skill whenever the user needs help with Agent, Browser, or Tools configuration, is writing code that imports from browseruse, asks about @sandbox deployment, supported LLM models, Actor API, custom tools, lifecycle…

browser-use/browser-use · 137 tokens

batch-processing-clinical-text

Run large-scale batch NER, PII extraction, or de-identification over many clinical notes on-device with OpenMed, with sharding, checkpointing, resumability, and append-only JSONL output. Use when the user needs to process a corpus or folder of notes, de-identify a dataset, run NER over thousands of documents, build a…

maziyarpanahi/openmed · 161 tokens

coding-hcc-risk-adjustment

Maps chronic conditions extracted by OpenMed to CMS-HCC V28 risk-adjustment categories and estimates a RAF (Risk Adjustment Factor) score as decision support. Use when the user wants to surface risk-adjustable diagnoses from notes, map ICD-10-CM codes to HCC categories, estimate or reconcile a patient/panel RAF, find…

maziyarpanahi/openmed · 200 tokens