Content
62%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 workflow is a model of explicit sequencing with genuine validation and recovery loops, and the in-file guidance is specific down to paths and tool parameters. It is dragged down by the two referenced bundle files (template.html, swe.md) being missing — the workflow's most executable assets — and by a body length that includes recurring justificatory padding.
Suggestions
Ship the referenced bundle files (template.html and swe.md) in the skill directory, or remove the references and inline the minimal required content (the state-block field list, a minimal tab skeleton) so steps 3 and the refresh section are self-sufficient.
Trim justificatory asides (the SVG benefits list, the hub/slug rationale, the coffee-chat framing) to their operative instruction — roughly: 'diagrams are inline SVG, not images; skip the dependency DAG; the slug is opaque so register on the hub if one exists'.
Tighten the multi-clause sentences in the tab catalog and Conventions sections into short normative statements, moving the 'why' to a single line per rule at most.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Mostly efficient operational content (paths, env vars, tool parameters, config schema) rather than concept explanations, but the ~250-line body has padded passages — long justificatory asides ('SVG keeps the page self-contained, scales crisply, wraps with the layout…', the hub-registration rationale, the 'what a colleague would tell you over coffee' framing) and very long multi-clause sentences that could be tightened without losing instruction. Not 4: the trimming opportunities are recurring, not minor; not 2: the padding is justification prose, not explanations of things Claude already knows. | 3 / 5 |
Actionability | The written guidance is unusually concrete (exact paths like ${CLAUDE_PLUGIN_DATA}/artifacts/<slug>/config.md, exact tool call parameters, an exact state-block JSON shape), but key executable details are delegated to files that do not exist in the skill directory: step 3 says 'Generate the HTML from template.html in this skill directory' and the refresh section defers the state-block field list to 'swe.md — don't improvise a different shape', yet neither file is present in the bundle. Not 4: missing core dependencies is more than a minor gap; not 2: the in-file guidance still specifies most of the work concretely. | 3 / 5 |
Workflow Clarity | The 8-step workflow is clearly sequenced with explicit validation checkpoints and feedback loops: step 4 re-reads the output for clipping/overflow before publishing and re-inspects after; step 8 handles publish conflicts ('WebFetch the URL… reconcile, then publish again'); the refresh section defines delta handling, missing-render recovery, and stale-data marking ('keep the previous values, mark exactly which rows or sections are stale, and never fill gaps from memory'). Not 4: validation is not merely present but looped with error-recovery paths. | 5 / 5 |
Progressive Disclosure | Good structure in principle — a domain-neutral overview with clearly signaled, one-level-deep references ('The SWE specifics… are in swe.md', 'Generate the HTML from template.html', a Files section listing both) — but scored against the actual bundle, both referenced paths are absent from the skill directory, so the disclosed content is unreachable and the SWE workflow and template the steps depend on cannot be loaded. Not 4: broken references are a real navigation failure, not an organization gap; not 2: the SKILL.md body itself is well sectioned and the reference structure is correct. | 3 / 5 |
Total | 14 / 20 Passed |