release: 0.5.0 — stop spending Tier A on Skills that work - #1
Merged
Merged
Conversation
The health report was the loudest part of this tool and the least trustworthy. On a real library of 213 Skills it raised 27 decisions; 11 of them were about files nobody loads or paths belonging to other people, and the one button on the screen could not act on any of them. - Underscored directory names load and trigger normally. Reporting them as broken put the highest severity on Skills that work. - The trash showed empty whenever one entry was a dangling link, because statSync followed the link and the throw emptied the whole listing. - Somebody else's home path arrives inside a third-party Skill and cannot usefully be edited locally, so it belongs in the background notes. - .bak and .orig copies are not loaded by any Agent; a path found in one is a note about a file nobody reads. Credential scanning still covers them. - Findings nothing can act on no longer claim to be fixable, and now carry the command to run instead. undo asks before reversing a batch. A run of describe/categorize/note inside 30 minutes shares one backup session, so a single undo could take back dozens of hand-written blurbs — and when that batch created overrides.json, reversing it removed the file with no copy to restore and no redo. It printed the count afterwards; it now prints it first and waits for --yes. The published package drops the local design workspace: files names web/index.html rather than all of web/, taking the tarball from 11.2 MB unpacked back to 315 kB. Also: a prerelease is superseded by its own release, a leaderboard missing its seed repositories expires in an hour rather than a week, both READMEs list the mutating commands and how to stop the dashboard, and SKILL.md finally tells an agent that remove/trash/link/unlink/update/backups/note exist. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
第二轮对抗式审查的结果,这次瞄准体检报告本身。
在一个 213 个 Skill 的真实库上,体检提出 27 项「需要你判断」,其中 11 项讲的是没人加载的文件或别人的路径,而屏幕上唯一的按钮对它们一项都动不了。修完是 16 项。
修了什么
cc_diary/tao_diary等在用的 Skill 加载触发都正常,规则的前提不成立statSync跟随软链抛错,catch 把整份清单清零;实测 1 条 → 0 条.bak/.orig副本被扫描fixable: true却没有任何入口能修false,建议里直接给命令isNewer("0.4.0-beta.1","0.4.0")曾为 false行为变更
undo在批量撤销前要求--yes。 30 分钟内连续的describe/categorize/note共用一个备份会话,一次undo可能退掉几十条手写介绍;如果那批还创建了overrides.json,反转会直接删掉文件,没有副本可恢复也没有 redo。原本数量是事后才打印的。files从web/收窄到web/index.html,解包体积 11.2 MB → 315 kB。报告界面重做
原本列表和按钮讲的是两件事却挤在一起(标题说 27 项、按钮说 2 条,中间没有分界)。现在拆成「可以一键处理」和「需要你自己处理」两段,并按规则归组——15 条同类折成一行,真正不同的那条才浮出来。
验证
npm test98/98 通过npm pack --dry-run:27 个文件,103.6 kB🤖 Generated with Claude Code