refactor(aidd-dev): lean the assert skill for small models
维护者通常 1 天内回复
还没有人认领这个 Issue。
评估
- 难度
- 4/5
- 预计耗时
- 3-5 天
- 新手友好度
- 55/100
- Issue 类型
- 重构
- 描述清晰度
- 基本清楚
- 活跃度
- 活跃
- 技术栈
- typescript
调研方向
Read the existing aidd-dev:03-assert skill files, then compare the brainstorm skill issue #1071, the plan skill, and aidd_docs/memory/coding-assertions.md. Restructure the skill around a Mermaid action router, recipe steps with an explicit loop, one Output per action, and action-owned rules without duplication. Run the pre-commit gates and compare the listed model scenarios; done means fewer words, green assertions at least as often at equal or lower cost, and stopping when assertions pass or need user input.
由索引模型根据 Issue 内容生成。
描述
Problem
aidd-dev:03-assert is 5 files and 979 words: three actions (assert, assert-architecture, assert-frontend) that iterate the project's coding assertions until they pass. Under opencode with Mistral small and medium models, every extra file read re-sends the whole context, and the iterate-until-green loop forces a re-read on each turn.
Scope
- Same method as the brainstorm and plan skills: mermaid router with action paths, recipe steps with the loop explicit, one
Outputper action, rules only where the action owns them, no duplicate between actions and references. - Check the router on opencode with
mistral-small-2603andmistral-medium-2604.
Acceptance criteria
- The pre-commit gates pass, with fewer words than today.
- On the same scenarios, Claude
haikuandsonnetandmistral-medium-2604reach green assertions at least as often as the original, at lower or equal cost. - The loop stops when the assertions pass or when a failure needs the user.
Prior art in this repo
- The brainstorm skill issue #1071 and the plan skill issue.
aidd_docs/memory/coding-assertions.md.
Out of scope
plan,implement,review: separate issues.- Changing the project's own assertions.
- 主要语言
- TypeScript
- 星标
- 481
- 派生
- 45
- 平均合并
- 1 天 6 小时
- 30 天内合并 PR
- 82
环境准备
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。
ai-driven-dev/framework 的其他 Issue
-
难度 2/5 1-3 小时 新手友好度 88/100
ai-driven-dev/framework#1002 ·
维护者通常 1 天内回复
-
feat(aidd-dev/test): enforce test effectiveness and test-double discipline可能已有人在做 关联的 PR 仍在进行中或已合并。 未关闭
难度 2/5 1-3 小时 新手友好度 82/100
ai-driven-dev/framework#952 ·
维护者通常 1 天内回复
-
refactor(aidd-orchestrator): the check zone says when to stop, and reviews its axes in one round未关闭
难度 2/5 1-3 小时 新手友好度 76/100
ai-driven-dev/framework#887 · 2 条评论 ·
维护者通常 1 天内回复
-
难度 4/5 3-5 天 新手友好度 45/100
ai-driven-dev/framework#1079 ·
维护者通常 1 天内回复
-
难度 4/5 3-5 天 新手友好度 48/100
ai-driven-dev/framework#1078 ·
维护者通常 1 天内回复
查看 ai-driven-dev/framework 的全部 Issue
相似的 Issue
-
clawsweeper:fix-shape-clear clawsweeper:queueable-fix clawsweeper:source-repro impact:ux-friction issue-rating: 🦞 diamond lobster no-stale P3
难度 2/5 1-3 小时 新手友好度 73/100
openclaw/openclaw#182077 · 1 条评论 · 1 个 reaction ·
维护者通常 1 天内回复
-
aikido high security
难度 2/5 1-3 小时 新手友好度 75/100
dequelabs/axe-linter-action#161 ·
-
[Feature Request] Update the DimCode agent icon to the 45:32 brand mark可能已有人在做 @eric8810 今天认领。 未关闭
难度 2/5 1-3 小时 新手友好度 72/100
维护者通常 1 天内回复
-
难度 2/5 1-3 小时 新手友好度 66/100
lyswhut/lx-music-desktop#3026 ·
-
难度 2/5 1-3 小时 新手友好度 76/100
abhigyanpatwari/GitNexus#3704 ·
维护者通常 1 天内回复