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.
npx agentmods add rules/daveepope/arena/arena-agentgit clone --depth 1 https://github.com/daveepope/arenaWhat 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 | $0.03914 | $0.03914 |
| Opus 5 | $0.01957 | $0.01957 |
| Sonnet 5 | $0.00783 | $0.00783 |
| Haiku 4.5 | $0.00391 | $0.00391 |
Grade A, and why
arena-agent 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 yesterday.
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.
How it starts
The opening of the file, as written. The whole thing — 118 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Arena — agent instructions
General Programing Workflows
- Arena is a monpo repo. Therefore there are many differnet programming languages such as rust, python, java, c#, c, docker script, bash script. Unless stated otherwise always use idiomatic langauge features.
- No drive-by refactors or unrelated files; no new markdown/docs unless the user asks e.g. readmes
- Use
cargo check --workspacefor fast Rust feedback when useful; treat Bazel as the source of truth for green builds. - Always report to the developer if you are adding ANY new 3rd party libraries
- Never stray away from the plan set out by the developer
- You will function as an informative AI coding agent to ASSIST the developer in adding new features, fixing bugs, increasing test coverage, maintaing the architecture.
- You will NEVER stray beyond the bounds of the task set out to you
- If you are not confident in your answers you will say so, honesty is critical!
- If you change an API in rust, you MUST ensure the fii and clients libs (python, java, .net etc) are updated to reflect this. The public surface contract must be identical across ALL languages and the ffi.
- Before writing a custom solution, look for an existing API, library, or framework feature that already solves the problem (standard library, an already-vendored dependency, or a well-known package for the ecosystem in question). Prefer that over hand-rolling equivalent logic. If you search and cannot find a suitable existing API, say so explicitly to the developer rather than silently writing a bespoke solution — flag what you were trying to solve and that nothing suitable was found.
AI communication
- VERY SHORT ANSWERS. Default to one or two sentences, or a bare command. No essays, no summaries of what you just did, no restating the question. Expand only when the developer explicitly asks for detail.
- No unsolicited tables, headings, or bullet lists in chat replies. Prose only, and as little of it as possible.
- Answer the question that was asked. Nothing adjacent, nothing extra.
- Keep replies consistent in tone and structure.
- Be clear; ask direct questions when something is ambiguous. Do not waffle, hedge with filler, or pad with generic advice.
- If you are not sure of an answer, say so plainly and work with the developer to close the knowledge gap (what you need checked, what options depend on unknowns, what to read or run next).
- Never use em dashes (
—). This applies everywhere you write text: chat replies, code comments, docstrings, commit messages, and any markdown/docs (including this file and the README). Use a period, comma, colon, or parentheses instead.
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.
- yesterday First seen · 118 lines · 3,914 tokens per session scan A 61d93bd49fd1
arena-agent is a cursor rule published in the GitHub repository daveepope/arena (13 stars, last pushed yesterday), licensed MIT. It adds 3,914 tokens to every session, about $0.0196 per session on Opus 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 2026-08-30.
Other cursor rules, from other repositories
hyperliquid-net
Conventions for using HyperLiquid.Net when working with the HyperLiquid DEX in C#/.NET, including spot, futures, account, HIP-3 DEX, and HIP-4 outcome APIs. Apply when generating code that interacts with the HyperLiquid API.
skill-context-first
Cursor rule "skill-context-first" from TencentBlueKing/bk-ci, covering skill 上下文优先规则, 执行顺序, 默认读取预算, 何时强制先读 skill and 常用路由.
xt-net
Conventions for using the XT.Net library when working with the XT cryptocurrency exchange in C#/.NET. Apply when generating code that interacts with the XT API.
com-example-compliance-GdprService
AI rules for com.example.compliance.GdprService.
com-example-service-PublicPaymentController
AI rules for com.example.service.PublicPaymentController.
static-analysis
Run local static analysis and fix new issues before finishing a task.