Content
78%Weight 40%Scale 1-5Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
The content is a strong, highly actionable iOS/macOS development reference with excellent progressive disclosure via verified bundle files. Its main weakness is redundancy between the Critical Warnings table and the detailed later sections, plus a few destructive commands that could use explicit validation steps.
Suggestions
Dedupe the Critical Warnings table against the SPM and Code Signing sections: keep the table as a quick index with anchor links instead of restating the EMFILE, teamId, adhoc-fallback, and @rpath details in two places.
Add an explicit validation or rollback step for destructive commands, e.g. verify `pkill -9 Xcode` succeeded and confirm the DerivedData path before `rm -rf`, and note how to undo the global git proxy setting.
Consider moving the large osxSign TypeScript block and the full code-signing secrets tables into references/apple-codesign-notarize.md (already linked) to tighten the SKILL.md overview further.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is well-structured with tables and code, but several issues are stated twice — the @rpath, EMFILE, teamId, and adhoc-fallback items appear both in the Critical Warnings table and again verbatim in the SPM and Code Signing sections — which is more than minor trimming. | 3 / 5 |
Actionability | Copy-paste-ready project.yml, concrete xcodebuild/security/codesign commands, and WRONG-vs-CORRECT Swift patterns directly cover the common cases with executable guidance. | 5 / 5 |
Workflow Clarity | Key fragile workflows have explicit checkpoints (the 5-step signing checklist, fail-fast three-layer defense, 'Verify signing' command, numbered SPM embed fix), but a few destructive commands such as `pkill -9 Xcode && rm -rf DerivedData` and global git proxy config lack explicit validation or rollback guidance. | 4 / 5 |
Progressive Disclosure | SKILL.md is a clear overview with well-signaled, one-level-deep links to five real reference files (all verified present), each labeled both inline and in a Resources section, with no nested references. | 5 / 5 |
Total | 17 / 20 Passed |