Content
67%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.
A highly actionable, well-referenced skill body with excellent executable commands, complete config examples, and a genuine one-level-deep reference bundle with a load-on-demand index. Its weaknesses are self-duplication (the Summary and When-to-Use sections repeat earlier content), a dated version section, and a few inaccurate or dangling pointers (.windsurfrules for Claude Code, README.md, dependency-upgrade skill).
Suggestions
Delete or drastically shrink the Summary section — it restates 'When to Use', the provider comparison, and key advantages already covered above — and drop the dated 'Package Versions (verified 2026-08-03)' section or move it to a maintenance/old-patterns note.
Fix the incorrect and dangling pointers: Claude Code rules should not be listed as `.windsurfrules` (that is Windsurf's file), and references to a nonexistent README.md and `dependency-upgrade` skill should be removed or replaced with real bundle paths.
Add inline failure branches to the main workflows (e.g., after 'Run `npx ultracite doctor`' in the migration steps, state what to do when doctor reports invalid preset paths), and trim inlined provider/migration detail in favor of the already excellent reference files.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The bulk is efficient, executable content (commands, configs, flags), but there is real padding: the ~55-line Summary section duplicates 'When to Use', the provider comparison, and the key advantages verbatim; install commands appear three times; and the 'Package Versions (verified 2026-08-03)' section embeds time-sensitive date/version info outside any old-patterns section, which the guidelines say should penalize conciseness. Not 2 because the document rarely explains concepts Claude already knows and most sections earn their tokens; not 4 because the duplication and dated version info go beyond 'minor instances'. | 3 / 5 |
Actionability | Nearly everything is copy-paste ready: `bun x ultracite init --linter biome`, full biome.jsonc heredocs, VS Code settings JSON, hook scripts, and documented CLI flags with expected `doctor` output. Not 5 because of concrete errors that would misdirect action: Claude Code rules are listed as generating `.windsurfrules` (that is Windsurf's file), and 'See README.md' plus 'Load the `dependency-upgrade` skill' point at files that do not exist in this bundle. | 4 / 5 |
Workflow Clarity | Sequences are clear and validation is mostly present: the v6→v7 migration steps include 'Run `npx ultracite doctor` to validate configuration' and 'Test linting: `npx ultracite check .`', and 'Verify Installation' shows expected command output. Not 5 because failure branches are not inline (no 'if validation fails, do X' loop in the workflow itself — recovery is deferred to the troubleshooting reference), and the init flow lists 10 steps without checkpoints between them. | 4 / 5 |
Progressive Disclosure | Scored against the actual bundle: all 15 markdown references and 2 scripts cited in the body exist, are one level deep, and are well signaled both inline ('For complete X, see references/X.md') and via a dedicated 'When to Load References' index mapping topics to files. Not 5 because the SKILL.md itself inlines substantial detail that duplicates the reference files (full provider sections, migration examples, monorepo structures), and the dangling README.md / dependency-upgrade pointers break navigation. | 4 / 5 |
Total | 15 / 20 Passed |