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 instructions/ai-twinkle/eval/claude-mdgit clone --depth 1 https://github.com/ai-twinkle/EvalWhat 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 | $0.12871 | $0.12871 |
| Opus 5 | $0.06435 | $0.06435 |
| Sonnet 5 | $0.02574 | $0.02574 |
| Haiku 4.5 | $0.01287 | $0.01287 |
Grade A, and why
Eval CLAUDE.md scanned grade A with 1 finding 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 2d 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.
Runs shell commandslowCapability
Expected in a hook, worth knowing in a rule or an instructions file.
- 以 subprocess、os.system 或任何方式執行 `vllm serve`、`ollama run`、`python -m ...` 等啟動模型服務的指令 How it starts
The opening of the file, as written. The whole thing — 858 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Twinkle Eval — 專案規範手冊(CLAUDE.md)
本文件是所有開發者與 coding agent 的強制閱讀文件。 在修改任何程式碼之前,必須先完整閱讀本文件,確保所有變更符合本專案的設計理念與規範。 若提議的變更與本文件有所衝突,必須先至 GitHub 開立 Issue 進行討論,而非直接提交 PR 要求合入。
目錄
- 專案定位與設計理念
- 核心設計原則(不得違反)
- 架構總覽
- 模組職責邊界
- 擴充規範——如何正確新增功能
- 新增評測 Benchmark 的完整規範
- 必須先開 Issue 的情況
- 程式碼風格規範
- 設定檔規範(config.yaml)
- 輸出格式規範
- 依賴管理規範
- CLI 設計規範
- 提交 PR 前的 Checklist
- 專案現況快照
- 貢獻者
1. 專案定位與設計理念
1.1 誕生背景
2025 年初,推理模型(reasoning model)開始大量出現。這類模型在輸出正式答案之前,會先產生大量的推理過程(chain-of-thought),導致每次 API 呼叫的回應時間遠高於傳統模型。
然而,當時現有的評測框架(如 iKala/ievals)採用同步、逐題呼叫的設計,面對推理模型時評測時間會等比例放大,一個完整的 benchmark 動輒耗費數小時。對於正在進行模型訓練迭代的團隊而言,這意味著每次訓練完都需要等待過久才能得到評測反饋,白白浪費 GPU 運行時間與開發週期。
Twinkle Eval 因此而生:以並行 API 請求為核心手段,讓評測速度不再是訓練迭代的瓶頸。實測相比 iKala/ievals 快 9–17 倍,使團隊能夠快速取得評測結果、驅動下一輪訓練決策。
1.2 核心設計哲學
「輕量、單機、即裝即用」是這個專案的根本設計方向。
本專案從一開始就以「pip install twinkle-eval 後,在單台機器上即可執行完整評測」為設計約束。這個約束是刻意的:
- 不依賴叢集基礎設施:不需要 SLURM、Kubernetes、或任何分散式排程系統才能運作
- 不需要特殊硬體:評測本身不需要 GPU,只需要能呼叫 API 的網路環境
- 降低使用門檻:任何人在任何環境(本機、Colab、CI/CD)都能直接執行
多節點分散式評測(如 SLURM 支援)若作為 PR 提交,定位是擴充功能(extension),不得影響單機執行路徑的正確性與簡潔性。核心程式碼必須在不依賴任何分散式元件的情況下完整運作。
1.3 關鍵設計邊界:本專案永遠不啟動模型端點
Twinkle Eval 不負責啟動、部署、或管理任何 LLM 服務。
本專案的職責範圍嚴格限定於:
「拿著評測題目,去呼叫已經在外部運行的 API 端點,取得回答,並計算評測指標。」
使用者需要自行在外部啟動模型服務(vLLM、Ollama、OpenAI、NVIDIA Build 等),再將端點的 base_url 填入 config.yaml,Twinkle Eval 才開始工作。
這個邊界意味著:
- 本專案程式碼中不得出現任何啟動
vllm serve、ollama run或其他模型服務的邏輯 - 本專案不管理模型的生命週期(啟動、關閉、重啟)
- 若 API 端點無回應,本專案的責任是報錯退出,而非嘗試修復或重啟服務
任何試圖在程式碼內部啟動模型服務的 PR,在未開 Issue 討論並取得 maintainer 明確同意前,不得合入。
1.4 核心目標
- 高效:以並行請求大幅縮短評測時間,讓評測不再是訓練迭代的瓶頸
- 客觀:透過選項隨機排列,消除模型對選項位置的偏好(參考 Changing Answer Order Can Decrease MMLU Accuracy)
- 穩定性量化:支援多次執行並計算標準差,反映模型一致性
- 易擴充:模組化設計,讓新增 LLM 後端、評測策略、輸出格式都不需修改核心邏輯
- API 相容優先:以 OpenAI 相容 API 為統一介面,不綁定特定模型或服務商
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.
- 2d ago First seen · 858 lines · 12,871 tokens per session scan A 153bd6fb555e
Eval CLAUDE.md is an instructions file published in the GitHub repository ai-twinkle/Eval (109 stars, last pushed 8d ago), licensed MIT. It adds 12,871 tokens to every session, about $0.0644 per session on Opus 5. A static security scan graded it A with 1 finding (runs shell commands). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other instructions, from other repositories
helicone AGENTS.md
AGENTS.md instructions for Helicone/helicone, covering repository guidelines, project structure & module organization, build, test, and development commands, coding style & naming conventions and testing guidelines.
ClawBench AGENTS.md
Instructions for TIGER-AI-Lab/ClawBench, covering clawbench -- agent context, what this is, project structure, setup and 2. configure at least one model.
eval-view AGENTS.md
Instructions for hidai25/eval-view, covering evalview agent instructions, what evalview is, core concepts, testcase and evaluationresult.
langium-ai CLAUDE.md
Instructions for eclipse-langium/langium-ai, covering claude.md, project overview, commands, build & test (all workspaces) and lint & format.
argus-validation-benchmarks CLAUDE.md
Instructions for pensar-x/argus-validation-benchmarks, covering project overview, what you're building, the goal, success criteria and what is apex?.
skill-optimizer AGENTS.md
Instructions for fastxyz/skill-optimizer, covering agents.md, project overview, key commands, important files and installation docs.