# Builderr starter brief: trading agent

## The job

Build a small agent that turns shared market data into portfolio decisions. The current challenge is a 60-day forward test against Arnav's verified winning agent.

This is paper trading. You do not connect a brokerage account, deposit money, or provide an API key.

## What the agent receives

- The same daily market bars and portfolio state as every other entry.
- A fixed set of rules for position limits, leverage, fills, and data access.

## What it must return

- Portfolio orders from one function in the starter template.
- A strategy that runs cleanly without lookahead or outside data.
- No position or leverage violations.

## What matters

1. Beat the previous round's winning agent on the forward live market window.
2. Survive bad weeks instead of making one oversized bet.
3. Follow the same limits and data rules as everyone else.

## Endpoint entries and winner verification

You may submit an HTTPS endpoint and keep its code private while the round is live. Endpoint leaderboard results remain provisional.

If you finish provisionally in the top 3, you must give Builderr read-only access to the exact agent version used during scoring. Builderr reproduces its behavior from saved point-in-time inputs and reviews it for lookahead and rule violations before issuing a prize, winner status, or certificate. If the exact agent is not provided or cannot be reproduced, the entry is prize-ineligible and the next eligible entry moves up.

You keep ownership. Confirmed top-3 agents are shared with the challenge backers under the published prize terms; lower-ranked agents are not shared.

## A beginner-friendly way to start

1. Open the [starter template](https://github.com/builderr-ai/builderr-trading-template).
2. Rename the baseline file as shown in the template and run `python preview.py`.
3. Ask an AI coding assistant to explain the example strategy and change one rule at a time.
4. Push the repository and email the URL to `submit@builderr.ai`.

You do not need a finance degree or an LLM. The template includes readable reference strategies. You do need to understand what your agent is doing and check that the preview passes before submitting.

## Full truth

The [challenge page](https://builderr.ai/trading-v0) contains the benchmark, score definition, revision cutoff, submission options, ownership terms, and safety limits. Read it before entering. This is a public paper-trading evaluation, not investment advice.