Content
67%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 well-structured, largely executable workflow guide with real one-level references and explicit verification steps. Its main weaknesses are repetition of the same image-form and copyright guidance across sections, the absence of the referenced scripts/ files from the bundle, and the inlined style-category pool that should live in its own reference file.
Suggestions
Deduplicate the repeated image-form list and copyright warnings — state them once (e.g. in the Aesthetic Contract) and reference that section elsewhere, cutting roughly 20 lines.
Move the BM category pool into a reference file (e.g. references/bm-style-categories.md) and keep only the selection rules in SKILL.md.
Add concrete commands for the changelog/weekly evidence-gathering step (e.g. git log / git diff --stat invocations) so those modes are as executable as PR mode.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The ~70-line BM category pool is project-specific knowledge Claude would not know and earns its place, but the image-form list ("poster, scene, tableau, cover, painting, photograph, staged artifact") is restated four to five times across PR Mode, Style Selection, the Aesthetic Contract, and Quality Bar, and the copyright and 'not a text-heavy infographic' warnings are each repeated. Mostly useful with several redundant restatements fits anchor 3, not the minor-trim level of 4. | 3 / 5 |
Actionability | PR mode supplies fully-specified executable commands (gh pr view with --jq, uv run with every flag, --print-prompt dry-run) and the theme/provenance marker blocks are concrete. Falls short of 5 because the cited scripts/generate_pr_infographic.py and scripts/generate_infographic.py are absent from the bundle, and changelog/weekly modes give no commands for assembling the diff evidence (e.g. no git diff invocation). | 4 / 5 |
Workflow Clarity | PR mode is clearly sequenced with a pre-spend local prompt test and a "Verify the image exists and is readable before reporting success" checkpoint, matching the clear-sequence-most-checkpoints anchor. Not 5 because changelog and weekly modes are unordered checklists with no numbered steps or explicit validation of the evidence pack before generation. | 4 / 5 |
Progressive Disclosure | Sections are well-organized, and both body references (references/prompt-blueprint.md, references/style-balance.md) exist in the bundle and are one level deep and clearly signaled ("Read these references before drafting the prompt"). Not 5 because the large BM category pool is inlined in SKILL.md — content that clearly belongs in a reference file, which the rubric treats as inlining rather than appropriate splitting. | 4 / 5 |
Total | 15 / 20 Passed |