Content
57%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
The body is well-organized with good progressive disclosure and concrete audit steps, but it leans on background stats and defers all executable code plus the post-optimization verification loop to the reference, leaving workflow and actionability at mid-level.
Suggestions
Move or trim the background performance statistics (HTTP-Archive percentages, mobile-conversion figures) so thresholds and steps aren't competing with context Claude already knows.
Add a short inline re-verification checkpoint after Fix (e.g., re-measure each optimized asset and confirm it is under threshold) so the batch workflow has an explicit feedback loop.
Include one small inline executable snippet (e.g., a single Sharp/Squoosh WebP command) in the body so guidance is copy-paste ready without opening the reference.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Mostly efficient and well-structured, but background stats (HTTP-Archive percentages, mobile-conversion figures) and threshold numbers duplicated between Quick Reference and Fix don't all earn their place. | 2 / 3 |
Actionability | Concrete numbered procedures and named tools with quality settings, but no executable inline code in the body — runnable detail is deferred entirely to the reference file. | 2 / 3 |
Workflow Clarity | Check/Fix are clearly sequenced, but the body prescribes batch optimization with no inline validate-after-compress/re-measure feedback loop, which caps workflow clarity for destructive batch operations. | 2 / 3 |
Progressive Disclosure | Clear overview with one well-signaled one-level-deep reference (references/rule.md, verified present) and cleanly organized sections, giving easy navigation. | 3 / 3 |
Total | 9 / 12 Passed |