- name
SKILL.md
技能檔案保留在發布者的儲存庫中。
在終端機中貼上複製的指令。它會將這個 SKILL.md 安裝到對應的代理中。
npx skills add MiniMax-AI/mcode-stack --skill harness-experiment -a cursor -y在終端機中貼上複製的指令。它會將這個 SKILL.md 安裝到對應的代理中。
npx skills add MiniMax-AI/mcode-stack --skill harness-experiment -a claude-code -y在終端機中貼上複製的指令。它會將這個 SKILL.md 安裝到對應的代理中。
npx skills add MiniMax-AI/mcode-stack --skill harness-experiment -a codex -y在終端機中貼上複製的指令。它會將這個 SKILL.md 安裝到對應的代理中。
npx skills add MiniMax-AI/mcode-stack --skill harness-experiment -a grok -y在 Grok Bot 聊天中傳送複製的訊息。Bot 會從 GitHub 取得這個 SKILL.md,並加入你所有 Bot 共用的技能中。
Install this skill: https://github.com/MiniMax-AI/mcode-stack/tree/HEAD/stacks/ronny/skills/harness-experiment來自發布者
Harness engineering moves forward by experiment. The principle: measure first, change one thing at a time, keep what survives testing.
Companion to MCode Engineering Notes 01, 02 and 05.
One prompt edit, one tool moved, one subagent merged, one threshold changed. If you need several, run several experiments.
For baseline and candidate, record:
Only report numbers you actually measured, with run counts.
trajectory-review). Find out why it flipped.Scaffolding built around a model's weak spots ages fastest. After each model upgrade, rerun the eval and ask which parts of the harness still earn their keep.
A short experiment record: hypothesis, the single change, eval set and run count, the metric table for baseline and candidate, flipped tasks with causes, and the decision (keep / revert / iterate).
Run a harness change as a controlled experiment, with a small end-to-end eval, one change at a time, repeated runs, and looking at flipped tasks, request count, tokens and cache hits, not just the ave
---name: harness-experimentdescription: Run a harness change as a controlled experiment, with a small end-to-end eval, one change at a time, repeated…---# stacks/ronny/skills/harness-experiment/SKILL.md總覽
Run a harness change as a controlled experiment, with a small end-to-end eval, one change at a time, repeated runs, and looking at flipped tasks, request count, tokens and cache hits, not just the average score. Use before and after changing prompts, tools, skills, subagent layout, compaction or other agent scaffolding, or when asked whether a change helped.
常見問題
從發布者下載 stacks/ronny/skills/harness-experiment/SKILL.md,或複製路徑後開啟儲存庫。PlugMCP 只收錄這項技能,不代管檔案。
發布者將它存放在 https://github.com/MiniMax-AI/mcode-stack。
Claude、Cursor、Codex 和 Grok Build 可以讀取你從發布者儲存庫加入的 SKILL.md。在聊天中傳送技能的 GitHub 資料夾連結,Grok Bot 就會安裝它。