Content
82%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-crafted review skill body: rules are concrete and project-specific rather than generic, findings output is precisely specified, and scope boundaries against the general code review are explicit. The main improvement opportunities are tightening the slight overlap between the effect-avoidance rules and making the review process sequence more explicit.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense, project-specific directive text with no explanation of concepts Claude already knows (no Svelte primer, no library boilerplate), e.g. "Derive values; do not synchronize copies. Use `$derived` for computed state." Minor trimming is possible: rules 2 and 3 overlap on `$effect` avoidance, and rules 4 and 6 carry wordy exception clauses, which keeps it just below the 5 anchor's "every token earns its place." | 4 / 5 |
Actionability | For an instruction-only review skill the guidance is fully actionable: named APIs (`$effect`, `$derived`, `untrack`, `onMount`), a concrete worked example ("handle a new notification in the notification-update handler, rather than watching notification state with `$effect`"), specific verification behaviors (Back/Forward/reload, wide and narrow viewport checks), and a defined findings format ("file and line, the rule, the concrete consequence"). Per the rubric's scoring note, absence of code is not penalized when instruction-only guidance is this concrete. | 5 / 5 |
Workflow Clarity | A clear sequence is present in prose — review scope, read the frontend instructions, trace changed behavior through callers, apply the rules, then "Report findings in severity order, followed by checks run and remaining gaps" — with an explicit completion gate ("Apply these rules before declaring the review complete"). It sits below the 5 anchor because the process sequence is embedded in prose rather than laid out as an explicit ordered workflow with checkpoints, though it is not a destructive/batch skill so no cap applies. | 4 / 5 |
Progressive Disclosure | The single-file body is well organized (intro, numbered Rules, Findings) with clearly signaled one-level-deep links to sibling skills ([chatto-code-review](../chatto-code-review/SKILL.md), [chatto-client-placement](../chatto-client-placement/SKILL.md)) and the project's AGENTS.md; no bundle files exist. This matches the 4 anchor's "good structure; references mostly clear; minor organization gaps" — the 12 rules are cohesive in one file but the intro paragraph and long rule list could be organized more tightly. | 4 / 5 |
Total | 17 / 20 Passed |