Content
65%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 well-structured, actionable review skill with excellent progressive disclosure — the body stays an overview and correctly delegates code examples to references/rule.md. The main costs are redundancy between the intro, Quick Reference, and Explain sections (which re-teach HTML Claude already knows), and a missing validation step in the codebase-wide Check→Fix workflow.
Suggestions
Merge the intro paragraph and the Explain section: they repeat the same facts (no <img> means nothing renders in IE11/unsupported formats; alt/width/height belong on <img>; sources are evaluated in order), and the Explain section largely covers HTML knowledge Claude already has — keep only what is specific to this rule.
Add an explicit validation checkpoint at the end of the Fix workflow (e.g. 'after fixing, re-run the Check list and confirm the selected source in DevTools → Network → Img') so the codebase-wide scan-and-fix loop closes; currently Fix ends at 'Show the corrected <picture> markup' with no verification.
Move the six inline numbered conditions in Check to a short bulleted checklist (or reference the Verification section of references/rule.md) so the scan criteria are easier to apply element-by-element than as one run-on sentence.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is mostly efficient but repeats the same facts across sections: the intro's "A `<picture>` without an `<img>` fallback renders nothing in browsers that don't support `<picture>` (IE11)" and "where `alt`, `width`, `height`, `loading`, and `fetchpriority` attributes live" are restated nearly verbatim in Explain ("The `<img>` element is the fallback for browsers that don't support `<picture>` (IE11)", "The alt attribute must be on the `<img>`"), and the Quick Reference bullet "`<source>` elements are evaluated top-to-bottom; the first match wins" duplicates Explain's "Browsers evaluate `<source>` elements in document order". The Explain section also largely covers standard HTML knowledge Claude already has, which the rubric penalizes. Not a 2 because each section is short and there is no generic filler. | 3 / 5 |
Actionability | Check gives six concrete, verifiable conditions ("Every `<picture>` has an `<img>` as its last child", "Each `<source>` has either a srcset attribute or a media attribute") and Fix gives concrete templates like `<img src="{fallback-url}" alt="{description}" width="{w}" height="{h}">` plus ordering rules. It stops short of 5 because the body contains no complete, copy-paste-ready markup example — the actual HTML examples are deferred to `references/rule.md` — leaving minor gaps in the executable guidance. | 4 / 5 |
Workflow Clarity | A rough sequence exists (Check → Fix → Explain → Code Review), but there is no validation checkpoint: Fix ends with "Show the corrected <picture> markup" without any verify step (e.g. re-run the check or confirm in DevTools). Because the workflow is a batch operation — "Scan all <picture> elements in this codebase" — the rubric's cap for missing validation in batch workflows applies, holding this at 3 even though the individual steps are well defined. | 3 / 5 |
Progressive Disclosure | The body is a well-organized overview (Quick Reference, Check, Fix, Explain, Code Review) with the implementation detail — code examples, use cases, and verification — appropriately split into a single clearly-signaled, one-level-deep reference: "see `references/rule.md`" for "full implementation details, code examples, and framework-specific guidance". That file exists, contains the promised detail, and itself references no further files, matching the 5 anchor. | 5 / 5 |
Total | 15 / 20 Passed |