Content
71%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.
The body is a lean, well-structured, self-contained reference whose rules are concrete and clearly scoped, with strong failure-mode framing. Its main gap is the absence of even one worked example — a composed question with its minimal state and typed answer — which would move the guidance from principled to directly imitable.
Suggestions
Add one short worked example: a question composed in code from named inputs, the minimal denatured state it ships, and the typed answer (including a no-match case) — this directly addresses the lowest-scoring dimension.
Tighten the intro paragraph's exception sentence about the `todo --auto` auto-scoped ranking exception, and merge the Rule 2 noul guidance into Rule 4 to remove overlap.
Consider a brief authoring/review checklist that sequences the four rules, so a reader applying them to a draft question has an explicit order of checks.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense with no filler and explains nothing Claude already knows; each rule states its do/don't in a few tight sentences ("the same inputs always compose the same question, and a change in phrasing shows up as a diff"). It is not a 5 because the intro's stuffed exception sentence ("A narrow exception is the `todo --auto` cycle's own candidate ranking, the auto-scoped ranking exception...") could be trimmed, and Rule 2's noul guidance overlaps Rule 4. Not a 3, since nothing is padding or unnecessary explanation. | 4 / 5 |
Actionability | There is concrete directive guidance — "Compose question text with the project's language's string formatting from named inputs — never hand-type a question into a prompt-shaped literal", example option strings ("none of these fit", "no listed option"), and "record it next to the question id in the caller's own output" — but the skill never shows a worked example of a composed question, a minimal state payload, or a typed answer, leaving key details of what good output actually looks like missing. Per the instruction-only scoring note the absence of code is not itself penalized, but the incompleteness matches anchor 3 rather than anchor 4's mostly-complete guidance. | 3 / 5 |
Workflow Clarity | The four rules are clearly organized, each with explicit do/don't framing and failure-mode reasoning, and the closing section names the failure mode ("Silent defaulting — treating an absent answer as either polarity... is the failure mode these rules exist to prevent"). This is a single-purpose reference skill with no destructive or batch operations, so no validation loop is required, and the simple-skill exception could apply; it falls short of 5 only because there is no authoring/review sequence or checklist that ties the rules into an applicable order. | 4 / 5 |
Progressive Disclosure | No bundle files exist (references/, scripts/, assets/ are all absent) and none are needed: the body is a self-contained ~60-line set of rules with clean, well-navigable section headers ("## Rule 1 — ..." through "## What an absence means") and no content that clearly belongs in a separate file. This matches the rubric's simple-skill guidance that a self-contained, well-organized skill with no need for external references scores 5 on progressive disclosure. | 5 / 5 |
Total | 16 / 20 Passed |