Content
63%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 lean, well-sectioned, and highly actionable for reading, QA, and image conversion, with an excellent required verification loop. Its main weaknesses are that the editing and create-from-scratch workflows depend on reference files (editing.md, pptxgenjs.md) that are absent from the bundle, and the bundled scripts (add_slide.py, clean.py) are never mentioned, leaving the core editing path without executable inline guidance.
Suggestions
Include editing.md and pptxgenjs.md in the bundle (or inline minimal executable command sequences for the unpack → edit → clean → pack workflow) so the two primary workflows are not dead references.
Reference the bundled scripts directly in the Editing Workflow section — e.g., 'python scripts/add_slide.py' and 'python scripts/clean.py' — instead of the vague 'manipulate slides → edit content → clean' one-liner.
Add an explicit validation checkpoint inside the editing sequence (e.g., re-pack only after verifying content with markitdown) so the workflow does not rely solely on post-hoc QA.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense and efficient — command blocks, palette/typography tables, and prescriptive bullets with no padding about what a .pptx file is or how libraries work. It sits at the score-4 anchor ('Efficient; minor instances of over-explanation that could be trimmed'), e.g., the redundant closing line 'If a .pptx file needs to be opened, created, or touched, use this skill' in the description-adjacent guidance and a few 'Don't' bullets that restate design conventions Claude already knows. It is below score 5 only because a few lines (e.g., 'This creates slide-01.jpg, slide-02.jpg, etc.') and some repeated design directives could be tightened further. | 4 / 5 |
Actionability | Reading and QA paths are fully executable ('python -m markitdown output.pptx', the grep placeholder check, 'soffice --headless --convert-to pdf', a copy-paste subagent prompt), but the two primary creation/editing workflows delegate to editing.md and pptxgenjs.md, which do not exist in the bundle, leaving only vague one-liners like 'Unpack → manipulate slides → edit content → clean → pack' with no inline commands. This matches the score-3 anchor ('Some concrete guidance but incomplete... missing key details'); it is not score 4 because the main editing workflow lacks executable steps in the body, and not score 2 because reading, QA, and image-conversion paths are fully concrete. | 3 / 5 |
Workflow Clarity | The QA section provides a strong required feedback loop — 'List issues found... Fix issues... Re-verify affected slides... Do not declare success until you've completed at least one fix-and-verify cycle' — plus grep validation and a visual inspection checklist, matching most of the score-5 anchor. It lands at score 4 ('Clear sequence with most checkpoints present; minor validation gaps') because the editing workflow itself ('Analyze template → Unpack → manipulate → clean → pack') is compressed with no per-step validation between unpack and pack, relying entirely on post-hoc QA. | 4 / 5 |
Progressive Disclosure | Structure is good on paper — a Quick Reference table routing to editing.md and pptxgenjs.md — but per the guideline to score against the actual bundle structure, both referenced files are missing from the bundle, so navigation to two of the three main workflows is broken (scripts/add_slide.py and scripts/clean.py exist but are never referenced). This fits the score-3 anchor ('Some structure but could be better organized; references present but not clearly signaled / content that should be separate is inline'); it is above score 2 because the routing itself is clearly signaled, and below score 4 because missing primary reference files are more than a 'minor organization gap.' | 3 / 5 |
Total | 14 / 20 Passed |