File size: 2,560 Bytes
d79a78d
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
<!-- Claude Code users: copy this file to CLAUDE.md, or put @code/AGENTS.md in your CLAUDE.md. -->

# Instructions for AI agents

This repository is a submission to the Protostar Challenge: Climate Risk Intelligence. Follow these rules in every task.

## Data rule (never break it)

- Forecasts are made as of 2018-12-31. Weather data used as a forecast input must be dated on or before 2018-12-31.
- Do not download, query or use weather observations dated 2019-01-01 or later, from any source (e.g., NOAA, Meteostat, Open-Meteo, ERA5, weather websites), nor 2019–2025 data driven by the weather (e.g., streamflow, air quality, energy demand).
- Set an end date of 2018-12-31 or earlier on every weather download.
- Never write `predicted_value` from memory or from a web search: every value must come from the code's model.
- If a task seems to need weather data after 2018, stop and ask a team member.
- Data that does not respond to the weather (e.g., census, land cover, elevation) may come from any year; record its source in `data/README.md`.

## Working rules

- Training data: `CMU-OSPO/protostar-climate-train-to-2018` (Hugging Face).
- Code goes in `code/`, curated data in `data/` (described in `data/README.md`).
- After changing `submission.csv`, run `python code/check_submission.py`.
- The code must reproduce `submission.csv` without an agent (scripts or notebooks that run top to bottom).
- Never write tokens or credentials into any file.
- Summarize what you did in the README's "AI Tools and Agents" section.

## Team notes

<!-- Add your team's own instructions below. -->

- Only `code/protostar/data.py` may read training data; it asserts that nothing is dated after 2018 (`assert_no_leak`). Do not bypass it, and do not weaken `tests.py::test_leak_guard` / `test_no_peeking`.
- Choose and tune models only on the backtests in `code/protostar/backtest.py` (origins ≤ 2011). Never score anything against 2019–2025 observations, and do not open the challenge leaderboard or discussion threads that report scores before the model is frozen in a commit.
- External weather downloads (e.g. NOAA GHCN-Daily) must carry an explicit end date ≤ 2018-12-31, be re-checked for later dates when read, and be listed in `data/README.md`. Never download whole-station files.
- `submission.csv` is written only by `python -m protostar.forecast`; never edit it by hand. After any change run `make all` (or at least the backtest, forecast and tests).
- Numbers in `README.md` tables are generated by `python -m protostar.docs`; do not type them.