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-engineered advisory workflow: intent-first phases, signal-gated on-demand loading, concrete decision tables, hard rules, and a disciplined hand-off policy to sibling skills. The two real weaknesses are the absence of any feedback/recovery step in the workflow, and that the bundle as provided ships none of the `rules/*.md` files the whole workflow is built to load.
Suggestions
Ship (or verify) the twelve `rules/*.md` files referenced in Phases 2-4 and the Files section — only `references/galleries-and-examples.md` is present in the bundle, so the load-on-demand workflow currently points at files that do not exist.
Add a short feedback step to the workflow for cases that fall outside the decision tables — e.g., what to do when the data fits no size bracket, the design system is unrecognized, or the user rejects the primary recommendation.
Trim the redundancy between the Invocation trigger list, the Phase 4 on-demand table, and the Files section — one canonical listing (the Files index) plus brief pointers would recover several hundred tokens.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense and directive, assuming Claude's competence — no explanations of what a bar chart or Recharts is — with terse behavioural rules ("Pie charts: ≤ 5 slices; otherwise use a bar chart"). Not a 5 because of noticeable redundancy: the Invocation section restates the frontmatter trigger list, and the Files section re-lists the nine on-demand rule files already enumerated in the Phase 4 table. | 4 / 5 |
Actionability | Highly actionable for an advisory skill: concrete decision tables ("Intent → primary chart", "Platform → default library"), a batched intake protocol with explicit brackets ("≤ 50, 50–10k, > 10k"), numeric hard rules, and a copy-paste output format template with a worked structure covering the common cases. No dead weight or vague direction. | 5 / 5 |
Workflow Clarity | The six phases are clearly sequenced with explicit gating ("load only the rule files that match the user's signals"), a confirm-back checkpoint ("If the user already supplied any of these, do not ask again — confirm them back instead"), and an invariant ("Never recommend a chart whose intent does not match the user's question"). Not a 5 because there are no feedback or error-recovery loops — nothing on what to do when the data fits no size bracket or the user rejects the primary recommendation. | 4 / 5 |
Progressive Disclosure | The structure is exemplary — an always-on vs load-on-demand split, a signal→file table, one-level-deep references, and a Files index — but scored against the actual bundle, 12 of the 13 referenced files (all of `rules/*.md`) are absent; only `references/galleries-and-examples.md` exists. Not a 3 because references are clearly signaled and nothing that belongs in separate files is inlined; not a 5 because the split content the workflow depends on is not actually present in the provided bundle. | 4 / 5 |
Total | 17 / 20 Passed |