standing: Only now can do nothing while the card claims done
维护者通常 1 天内回复
还没有人认领这个 Issue。
- #1612 来自 @santoshkumarradha —— 已关闭，未合并
评估
调研方向
Reproduce the interactive CodeAF chat using the inventory CSV and markdown shopping plan, then trace the approval card through the stand result, execution path, and receipt. Verify that a one-time approval carries the requested action without persistence, that the receipt waits for execution evidence, and that decline and repeating approval retain their stated behavior.
由索引模型根据 Issue 内容生成。
描述
Defect
Choosing Only now, don't repeat on a repeating standing-task proposal immediately renders done now, nothing kept, even when no work has run. In live DeepSeek v4.1 Flash acceptance, the model then interpreted the answer as a decline and performed no tool call.
Replication
- Start an interactive CodeAF chat in a project containing an inventory CSV and a markdown shopping plan.
- Ask for a weekly task to read the inventory and refresh the plan without changing inventory or purchasing.
- On the approval card select Only now, don't repeat.
- The UI claims the work is done immediately. The stand result instructs the model to do an ordinary step but also says the person chose not to repeat; the recorded assistant response said nothing on disk changed and made no work calls.
Observed on #1596 head a955db54e. Original recording, terminal snapshot, transcript and receipt evidence are preserved on Spark. The live tool response was followed directly by a final reply, with zero work calls between them. A subsequent explicit user correction did recover the workflow.
Acceptance
- Approval for one execution must carry the actual requested action into the execution path, with no repeating arrangement installed.
- The approval receipt must distinguish authorization from completed execution; do not claim done until there is execution evidence.
- A real inventory-plan workflow must perform the once-approved refresh without the person having to repeat the instruction.
- Decline still performs nothing, and normal repeating approval still persists the repeating item.
Related to #1520 prompt/documentation review and discovered during sustained acceptance. Keep the fix separate for coordinated consolidation.
- 主要语言
- Go
- 星标
- 115
- 派生
- 14
- 平均合并
- 12 小时 30 分钟
- 30 天内合并 PR
- 323
环境准备
我们还没有检查这个项目的环境配置文件。先看它的 README，通用步骤见我们的新手贡献指南。
从这里开始
- 先读完整个 Issue，再读项目的贡献指南。
- 在 Issue 下留言说明你要接手 —— 这能避免两个人做同样的事。
- Fork 仓库，在一个分支上完成修改。
- 提交 Pull Request，并在描述里引用这个 Issue 编号。
Agent-Field/CodeAF 的其他 Issue
-
难度 2/5 1-3 小时 新手友好度 68/100
Agent-Field/CodeAF#1805 ·
维护者通常 1 天内回复
-
area:session bug sev:papercut
难度 2/5 1-3 小时 新手友好度 82/100
Agent-Field/CodeAF#1779 ·
维护者通常 1 天内回复
-
area:session bug sev:papercut
难度 2/5 1-3 小时 新手友好度 78/100
Agent-Field/CodeAF#1768 ·
维护者通常 1 天内回复
-
area:chat bug sev:papercut
难度 2/5 1-3 小时 新手友好度 78/100
Agent-Field/CodeAF#1742 ·
维护者通常 1 天内回复
-
area:build bug sev:papercut
难度 2/5 1-3 小时 新手友好度 88/100
Agent-Field/CodeAF#1679 ·
维护者通常 1 天内回复
查看 Agent-Field/CodeAF 的全部 Issue
相似的 Issue
-
darwin zsh
难度 1/5 1 小时以内 新手友好度 90/100
carapace-sh/carapace-bin#3846 ·
维护者通常 1 天内回复
-
难度 1/5 1 小时以内 新手友好度 70/100
golang-app/ecommerce#179 ·
维护者通常 1 天内回复
-
Simulator process crashes (nil pointer panic) when the dataset returns an error可能已有人在做 @EugeneVon 今天认领。 未关闭
难度 2/5 1-3 小时 新手友好度 78/100
llm-d/llm-d-inference-sim#772 ·
维护者通常 1 天内回复
-
refactor: decode numchildren with a lenient int type instead of a hand-written xProp.UnmarshalXML未关闭
难度 2/5 1-3 小时 新手友好度 84/100
crazy-goat/xdbg#125 ·
维护者通常 1 天内回复
-
150 points ci complexity:medium drips-wave help wanted medium
难度 2/5 1-3 小时 新手友好度 68/100
sorotrail/SoroBeacon#449 ·
维护者通常 1 天内回复