An agent that builds a Kaggle or MLE-Bench competition submission by training models, trying controlled changes, and writing a submission.csv file.

Use it to read prior competition notes, create a baseline, test one change at a time, validate the output file, and hand it to a grader.

It provides a repeatable process for moving from a first model to a checked submission while recording experiments.

builder — Agent for Claude Code by topprismdata/cultivating-ml-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.

Clone the repo

git clone --depth 1 https://github.com/topprismdata/cultivating-ml-agent

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.

README.md

[![agentmods](https://agentmods.dev/badge/agents/topprismdata/cultivating-ml-agent/builder/github.svg)](https://agentmods.dev/agents/topprismdata/cultivating-ml-agent/builder)

Your own site

<a href="https://agentmods.dev/agents/topprismdata/cultivating-ml-agent/builder"><img src="https://agentmods.dev/badge/agents/topprismdata/cultivating-ml-agent/builder/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/agents/topprismdata/cultivating-ml-agent/builder"><img src="https://agentmods.dev/badge/agents/topprismdata/cultivating-ml-agent/builder.svg" alt="Reviewed on agentmods" width="80" height="20"></a>

Per session 22 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.

When invoked 445 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	$0.00022	$0.00445
Opus 5.5	$0.00009	$0.00178
Sonnet 5.5	$0.00004	$0.00089
Haiku 4.5	$0.00002	$0.00044

Measured 2mo ago against content hash 67963f01260d, method: parsed. Prices are Anthropic first-party input rates as of 2026-10-07, from the pricing page.

Security

Grade A, and why

builder 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.

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.

ml-agent-code-template/.claude/agents/builder.md · 53 lines

What it actually says

Builder Agent

Role: Produces submissions/<competition>/submission.csv from a competition dataset. Stage: Hypothesis → Baseline → Iterate → Output Output: Always writes to submissions/<competition>/submission.csv Communicates with: Grader agent (validates the submission it produces)

Inputs

Competition slug (e.g., tps-may-2022, jigsaw-toxic-comment-classification-challenge)
Data path (default: ~/Library/Caches/mle-bench/data/<competition>/prepared/public/)
Submission path: submissions/<competition>/submission.csv

Responsibilities

Read memory first: Check memory/competitions/<similar>.md and memory/feedback_no_recheck_confirmed_dead.md
Build baseline: One simple model, get it producing output (don't optimize yet)
Iterate per SOP: One variable per experiment, log to memory/experiments/
Validate before output:
- File exists and is non-empty
- Header is correct
- All rows have values
- Submission size is reasonable
Hand off to Grader: After producing submission.csv, defer to grader for validation

Forbidden Actions

❌ Don't run mlebench grade directly — let Grader do it
❌ Don't submit to leaderboard unless explicitly asked
❌ Don't burn compute on hyperparameter tuning at stacking ceiling
❌ Don't skip the experiment logging step

Communication Protocol

When finished, emit a structured handoff:

=== HANDOFF TO GRADER ===
Competition: <slug>
Submission: <path>
OOF Score: <metric>
OOF vs Baseline: <delta>
Hypothesis tested: <one sentence>
Risks/caveats: <any>
=== END HANDOFF ===

The Grader agent takes this and runs validation.

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 · 53 lines · 22 tokens per session scan A 67963f01260d

Subscribe to this mod's changes

builder is an agent published in the GitHub repository topprismdata/cultivating-ml-agent (5 stars, last pushed 25d ago), licensed MIT. It adds 22 tokens to every session and 445 once invoked, about $0.0001 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.

Data and AI MLOps agent-memory agent-skills ai-agents knowledge-crystallization ml-experiments

builder next to 6 agents from other repositories
Mod	Type	Stars	Last push	Licence	Scan	Tokens	Works with
buildertopprismdata/cultivating-ml-agent	Agent	5	25d ago	MIT	A	22	Claude Code
cortexjeremylongshore/tons-of-skills-marketplace	Agent	2.9k	today	MIT	A	75	Claude Code
mlops-engineerpjt222/agent-almanac	Agent	43	yesterday	MIT	A	31	Claude Code
research-assistantSpillwaveSolutions/agent-brain	Agent	119	1mo ago	MIT	A	19	Claude Code
Demostheneswrg32786/aigent-os	Agent	24	today	MIT	A	69	Claude Code
mlops-engineerNickCrew/Claude-Cortex	Agent	66	4mo ago	MIT	A	53	Claude Code
ml-engineerNickCrew/Claude-Cortex	Agent	66	4mo ago	MIT	A	40	Claude Code