Borrowing it
Nothing to install: this file belongs to tim101010101/arena. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/tim101010101/arena/main/.claude/skills/arena-eval/SKILL.mdgit clone --depth 1 https://github.com/tim101010101/arenaWrote 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.
[](https://agentmods.dev/skills/tim101010101/arena/arena-eval)<a href="https://agentmods.dev/skills/tim101010101/arena/arena-eval"><img src="https://agentmods.dev/badge/skills/tim101010101/arena/arena-eval.svg" alt="Measured on agentmods" height="20"></a>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 | $0.00075 | $0.02136 |
| Opus 5 | $0.00037 | $0.01068 |
| Sonnet 5 | $0.00015 | $0.00427 |
| Haiku 4.5 | $0.00007 | $0.00214 |
Grade A, and why
arena-eval 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 6d 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.
How it starts
The opening of the file, as written. The whole thing — 205 lines — stays where its author put it; the contents beside it link to each section on GitHub.
arena-eval
Local-only evaluation harness for the arena project. The host agent (you, when triggered) orchestrates the run by calling the eval CLI for planning and aggregation, and dispatches parallel subagents to execute and judge cases.
When to use
Trigger when the user asks to evaluate, benchmark, or score the arena project — phrases like "eval arena", "run arena evals", "score the latest arena changes", "regression-test arena".
Do NOT trigger for the unit/integration test suite (bun test); those are different.
Inputs you accept from the user
- Tag filter: "eval the open ones", "only run code cases", "skip smoke" → translate to
--tags <list>/--exclude-tags <list>flags on the planner. - Models filter: "only use claude", "skip codex", "claude vs openai" → translate to
--only-models <list>on the runner (passed in each subagent'sbun evals/runner run). Default: empty (use every healthy adapter). - Run id: optional. Default: planner generates one.
- Repeats: optional. Default: 1. Bump only if the user asks for variance ("run each 3 times").
If the user gives no qualifier, run the full manifest at default settings.
Procedure
Follow these steps in order. Do not skip the planner — it generates the run id and reports directory that subagents must write into.
1. Pre-flight
Run a health check first. If no adapters are healthy, abort with a clear message:
bun src/index.ts health
If the JSON shows zero ok: true adapters, stop and tell the user no agent CLIs are
available; suggest installing claude / codex / etc.
2. Plan
Call the planner. It generates the run id, creates the reports directory, and returns the shard layout you will fan out:
bun evals/runner plan [--tags ...] [--exclude-tags ...]
Capture the JSON output. You will use:
run_idjudge_rubric_pathjudge_rubric_hashreports_dirshards[]— each shard is a list of cases for one subagent
3. Read the rubric once
Read the judge rubric file at judge_rubric_path so you can pass its full content to each
subagent. Do not paraphrase it. Subagents must judge against the literal rubric.
What ships with it
1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- 6d ago First seen · 205 lines · 75 tokens per session scan A aae65ffc8a7f
arena-eval is a skill published in the GitHub repository tim101010101/arena (1 stars, last pushed 3mo ago), licensed MIT. It adds 75 tokens to every session and 2,136 once invoked, about $0.0004 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 2026-08-31.
Other skills, from other repositories
testing-skills
Use when creating or editing Hyperstack skills, before shipping them, to verify they actually work under pressure and resist rationalization. Treats skill content as process-under-test with RED-GREEN-REFACTOR cycle enforced via subagent pressure scenarios.
behaviour-analysis
Systematic UI/UX behaviour analysis for interactive applications. Audits every user action, state transition, view mode, and edge case like an experienced QA + UX engineer. Produces a complete interaction matrix with expected vs actual behaviour, finds inconsistencies, dead states, and missing feedback. Use when…
debug-discipline
Use when encountering any bug, test failure, or unexpected behaviour. Root cause investigation is mandatory before any fix attempt.
test-first
Use when implementing any feature, bug fix, or behaviour change - before writing implementation code. Enforces test-before-code discipline.
research-engineer
An uncompromising Academic Research Engineer. Operates with absolute scientific rigor, objective criticism, and zero flair. Focuses on theoretical correctness, formal verification, and optimal implementation across any required technology.
tika-eval-compare
Compare extracts from two Tika builds over a corpus to detect regressions in content, encoding, exceptions, and embedded-document handling. Use for "compare before/after extracts", "eval this change against the corpus".