CLAUDE.md@benchmarks/arc-agi-3A
CLAUDE.md@benchmarks/arc-agi-3 from bitterbot-ai/bitterbot-desktop is a CLAUDE.md file: the standing instructions Claude Code reads in that repo.
mdr add bitterbot-ai/bitterbot-desktop/benchmarks/arc-agi-3/CLAUDE.md@git:20261012.b686058Pins git:20261012.b686058 in mdr.lock. Pin the fingerprint instead.
- Audit grade
- A, 14 of 14 checks passed
- Kind
- CLAUDE.md file
- Repository
- bitterbot-ai/bitterbot-desktop
- Path
benchmarks/arc-agi-3/CLAUDE.md- Versions recorded
- 3
- Latest version
- git:20261012.b686058, committed Oct 12, 2026
- First version recorded
- May 18, 2026
- Size
- 10,452 bytes
- Repository stars
- 2,480
- License
- MIT
Collected from public GitHub. Every version is saved exactly as it was and never changes. Install it at an exact version with the mdr tool, and every file is checked before it reaches your agent. The main file is checked against the SHA-256 fingerprint recorded here, and the others against the fingerprints GitHub keeps for its source commit. The automatic checks below grade the latest version, and the same checks always give the same file the same grade.
What the file says
You are a bot solving the ARC-AGI-3 benchmark. This file teaches you how to play. Read it at the start of every game. # What ARC-AGI-3 is ARC-AGI-3 is an interactive turn-based reasoning benchmark. Each game is a small grid-based world with rules you must discover by interacting with it. No instructions, no stated goals — figure out what's going on, infer the objective, and play efficiently. ## Core mechanics - **Observations** are JSON frames with one or more 2D grids (up to 64×64 cells, values 0..15). The last frame is the current state. Multi-frame responses are animations between actions. - **Actions** are: `RESET`, `ACTION1..7`. The first four are directional (1=up, 2=down, 3=left, 4=right). `ACTION5` is contextual (select / rotate / execute — varies per game). `ACTION6` clicks at `(x, y)` coordinates (0..63). `ACTION7` is undo. - **Game states**: `NOT_STARTED → NOT_FINISHED → WIN | GAME_OVER`. After GAME_OVER, only RESET is accepted. - **Scoring**: `level_score = (human_actions / ai_actions)^2`, capped at 1.15×. **Fewer actions = quadratically better score.** Reasoning + tool calls DON'T count — only commands submitted to the game count. …
Read the whole file at its exact version.
How to install
mdr add bitterbot-ai/bitterbot-desktop/benchmarks/arc-agi-3/CLAUDE.md@sha256:0de4a2c02c0bd45bThe line at the top of this page pins the label git:20261012.b686058, which is what most people want. This one pins the sha256 fingerprint of the contents instead, so it survives a label being moved. Either way the exact version is written to mdr.lock, and mdr install downloads and checks those contents again, so it installs them exactly or stops with an error.
[](https://markdownregistry.com/a/art_l5hnubwozcnmd32a)
3 badge image fetches in 30 days. What the badge shows.
Versions
The registry has recorded 3 versions of this file, between May 18, 2026 and Oct 12, 2026, about one every 74 days.
| version | committed | commit | size | audit | |
|---|---|---|---|---|---|
| git:20261012.b686058 latest | 2026-10-12 | b686058 | 10,452 B | A | view · diff |
| git:20260519.5e0d211 | 2026-05-19 | 5e0d211 | 10,452 B | A | view · diff |
| git:20260518.02a5b64 | 2026-05-18 | 02a5b64 | 9,961 B | A | view |
Audit of the latest version
- pass: Size between 200 bytes and 200 KB (10452 bytes)
- pass: No zero-width or bidi control characters
- pass: No instruction hidden inside an HTML comment
- pass: No link to an exfiltration or paste host
- pass: No credential-shaped string
- pass: No instruction to send local credentials anywhere
- pass: No text hidden with inline styles
- pass: No prompt-injection phrasing
- pass: No curl or wget piped into a shell
- pass: No recursive delete of root, home or parent
- pass: No instruction to read or print local credentials
- pass: No base64 blob over 200 characters
- pass: No link to a raw IP address
- pass: No script tag
Source
bitterbot-ai/bitterbot-desktop · 2,480 stars · license MIT · pushed 2026-11-03 · branch main
API
GET https://markdownregistry.com/api/v1/artifacts/art_l5hnubwozcnmd32a GET https://markdownregistry.com/api/v1/resolve?ref=bitterbot-ai/bitterbot-desktop/benchmarks/arc-agi-3/CLAUDE.md GET https://markdownregistry.com/api/v1/blob/0de4a2c02c0bd45b60cde5670c40f68c1f4b53ec168ea9b73ed04cf1bde8b985
Your AI assistant does the legwork. You hear about the deals worth your word. Hand yours the standing instructions at modelranch.com and it joins the network that reads files like this one.
Guides: what a CLAUDE.md file is · keeping AGENTS.md and CLAUDE.md in sync · which agent file does what.