繁體中文

問題修復

自動程式碼審查修回正常,先前改壞的提示詞撤回了

如果你的自動程式碼審查突然變得不對勁,這次把它修回原本的樣子了。先前一次更新不小心把 Codex 的 Guardian 自動審查功能(Guardian,像個會自動幫你檢查程式碼的 AI 審查員)改壞了——給它的提示詞(prompting,告訴 AI 該怎麼做事的指令措辭)調整後反而讓審查品質退步(regression,新版反而比舊版差的退步)。這次直接回滾(revert,把改動整個撤回、回到之前的版本),把舊版的審查政策、請求格式(request format,AI 發送需求時的訊息結構)和工具行為(tool behavior,AI 呼叫外部工具時的表現)全部恢復原狀。就像餐廳主廚試了新食譜結果常客都說變難吃,最後趕緊翻回舊食譜照原本的方式做菜。

English

BUGFIX

Automatic code review is back to normal after reverting the prompt change that broke it

If your automatic code review suddenly started acting up, this release puts it back the way it was. A prior change to the prompting (the wording of the instructions telling the AI how to do its job) for Codex's Guardian auto-review (Guardian, an AI reviewer that automatically inspects your code) caused a regression (a step backward where the new version is worse than the old one). This release reverts it (undoing a change and returning to the previous version), restoring the old review policy, request format (the message structure used when the AI sends a request), and tool behavior (how the AI acts when calling external tools). Like a restaurant chef who tried a new recipe, got panned by regulars for worse food, and hurriedly dug out the old recipe to cook the original way again.

原始 CHANGELOG

Original changelog自動程式碼審查修回正常,先前改壞的提示詞撤回了
Restored the previous Guardian auto-review policy, request format, and tool behavior after rolling back a prompting regression. (#32672)
#32672 [release/0.144] Revert "Update auto review prompting" @dylan-hurd-oai