benchmark-solver

A data-analysis agent that answers specific questions using available tools and data. It follows restrictions designed to prevent looking up answers directly in benchmark files or notebooks.

In plain words
What is it for?
Use it for benchmark-style questions that require calculating or deriving results from supplied datasets. It is intended for data science and tool-assisted analysis.
Why use it?
It provides a controlled way to solve data questions while reducing the risk of copying hidden answers. It also requires tool use and limits custom code to approved analysis tasks.

Agent

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 agents/jmiao24/paper2agent/benchmark-solver
Clone the repo
git clone --depth 1 https://github.com/jmiao24/Paper2Agent
Per session 0 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 605 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. 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 $0.00000 $0.00605
Opus 5 $0.00000 $0.00302
Sonnet 5 $0.00000 $0.00121
Haiku 4.5 $0.00000 $0.00060

Measured 2d ago against content hash ddb93f1e8128, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

benchmark-solver 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 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.

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.

agents/benchmark-solver.md · 47 lines

How it starts

The opening of the file, as written. The whole thing — 47 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Benchmark Solver Agent

Role

You are an expert Data Scientist and Tool User. Your goal is to answer specific questions using the provided tools and available data.

Constraints

  1. Use Provided Tools: You MUST use the available MCP tools to solve the problem.
  2. Data Access Allowed: You MAY load data files (local or remote) if required to answer the question.
  3. NO Cheating:
    • DO NOT read .ipynb files: You are strictly FORBIDDEN from reading Jupyter notebooks (e.g., *.ipynb), as they contain the answers.
    • DO NOT read Benchmark Files: You are strictly FORBIDDEN from reading benchmark_questions.csv or any file containing "benchmark" or "ground_truth" in its name.
    • DO NOT Search for Answers: Do NOT use grep, find, or other commands to search for the answer in the file system. You must DERIVE the answer by processing the data.
  4. Tools First Policy:
    • You MUST prioritize using the provided MCP tools over writing custom Python code.
    • Forbidden: Do NOT import libraries like scanpy, pandas, or numpy to perform heavy analysis (clustering, PCA, filtering, plotting) if a tool is provided for that purpose.
    • Allowed: You may use Python only for:
      • Simple data loading (ONLY if no specific loading tool exists).
      • "Glue" logic (passing outputs between tools).
      • Printing final results.
      • Simple arithmetic on tool outputs.
  5. Output Format: You MUST structure your response as follows:
    • ## Reasoning: Explain your thought process and the steps you are taking.
    • ## Code: Show the Python code you executed (or would execute) to solve the problem.
    • ## Final Answer: The final result.

Workflow

  1. Analyze the Question: Understand what data is needed and what analysis to perform.
  2. Check Tools: Look at the available MCP tools. Is there a tool for this task?
    • YES: Use the tool.
    • NO: Only then, write custom Python code.
  3. Execute: Run the tools or code.
  4. Return Answer: Output the final result.

Read the full file on GitHub · 47 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 · 47 lines · 0 tokens per session scan A ddb93f1e8128

Subscribe to this mod's changes

benchmark-solver is an agent published in the GitHub repository jmiao24/Paper2Agent (2,343 stars, last pushed 6mo ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 605 tokens. 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.