Content
56%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 skill body is a well-sequenced, information-dense workflow with genuinely useful non-obvious rules (viewport invariants, density limits, CSS negation gotcha). Its critical flaw is that the bundle is incomplete: every design/CSS reference file the workflow depends on is missing, making the core generation phases unexecutable as shipped, and the viewport rules are redundantly repeated across sections.
Suggestions
Ship the four missing referenced files (STYLE_PRESETS.md, viewport-base.css, html-template.md, animation-patterns.md) or inline the mandatory viewport CSS into SKILL.md so Phase 3 is executable.
Consolidate the viewport-fitting rules into one canonical section and reference it from Mode C / Phase 3 instead of restating them four times.
Add an explicit post-generation validation step (e.g., verify every slide fits 1280x720 with no overflow) and error-recovery guidance for failed pptx extraction in Phase 4.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is mostly dense with real, non-obvious guidance (density-limit table, breakpoints, the calc(-1 * clamp()) gotcha), but viewport rules are restated in four places (Core Principles, Viewport Fitting Rules, Mode C rules, Phase 3 key requirements) and the aesthetics lists repeat themselves ('Avoid generic fonts like Arial and Inter' then 'Overused font families (Inter, Roboto, Arial...)'). This is more tightening work than the 'minor instances' of a 4, but not the pervasive padding of a 2. | 3 / 5 |
Actionability | Concrete, executable guidance is strong in places (exact CSS declarations, per-slide density limits, the extraction command 'python scripts/extract-pptx.py <input.pptx> <output_dir>', AskUserQuestion specs), but key details are missing: the four files it instructs Claude to read and include (viewport-base.css, html-template.md, animation-patterns.md, STYLE_PRESETS.md) do not exist in the bundle, so Phase 2/3 cannot be executed as written. This lands on the 'incomplete; missing key details' anchor rather than 4's 'minor gaps'. | 3 / 5 |
Workflow Clarity | Phases 0-5 form a clear, gated sequence with user-confirmation checkpoints (Step 1.2 outline confirmation, Step 2.3 style pick, Phase 4 step 2) and Mode C has explicit post-modification verification ('After ANY modification, verify: ... content fits at 1280x720'). A 5 would require explicit validation after full generation (viewport-fit check for new decks) and error-recovery guidance for failed pptx extraction; not a 3 since checkpoints are present, not merely implied. | 4 / 5 |
Progressive Disclosure | The in-body structure is good — a 'Supporting Files' table with what/when-to-read and clearly signaled one-level-deep references ('Before generating, read these supporting files'). However, scored against the actual bundle, four of the five referenced files are absent (only scripts/extract-pptx.py exists), so the disclosure chain breaks at its first hop. This falls between 'references present but not clearly signaled' (3) and 'references mostly clear, minor gaps' (4) — the signaling is a 4-5 but the missing files are a major, not minor, gap, so 3 fits best. | 3 / 5 |
Total | 13 / 20 Passed |