Content
60%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 well organized with an effective Quick Reference router, dense concrete design guidance, and an outstanding QA verification loop, and the reading/QA paths are fully executable. However, the two primary workflows (editing and creating from scratch) delegate to editing.md and pptxgenjs.md, which are missing from the bundle, leaving those paths hollow, and the bundled validate.py is never surfaced as a pre-pack checkpoint.
Suggestions
Ship editing.md and pptxgenjs.md in the bundle (or inline their essential steps and commands in SKILL.md) — both are referenced by the Quick Reference but neither file exists, breaking the two primary task routes.
Add an explicit validation checkpoint to the editing workflow (e.g., `python scripts/office/validate.py unpacked/` before packing), since the script exists in the bundle but is never referenced, and show the pack/clean commands alongside unpack.
Trim editorial asides in the Design Ideas section ('won't impress anyone,' 'hallmark of AI-generated slides') and consider moving the palette/font tables to a design reference file to reduce main-file tokens.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Mostly lean, directive content — hex palettes, font pairings, pt sizes, and exact commands with little padding — but a few editorial asides ('Plain bullets on a white background won't impress anyone,' 'these are a hallmark of AI-generated slides') and justifying prose around the subagent advice could be trimmed, keeping it below the every-token-earns-its-place anchor. | 4 / 5 |
Actionability | The read/QA/conversion paths are copy-paste executable (markitdown, thumbnail.py, soffice.py, pdftoppm, the subagent inspection prompt), but the edit and create-from-scratch paths delegate entirely to editing.md and pptxgenjs.md, which do not exist in the bundle — the inline edit steps are arrow shorthand ('Unpack → manipulate slides → edit content → clean → pack') with no commands, and the create path has no actionable content at all. More than minor gaps, so below anchor 4. | 3 / 5 |
Workflow Clarity | The QA verification loop is exemplary — explicit fix-and-re-verify cycle, 'Do not declare success until you've completed at least one fix-and-verify cycle,' and a detailed inspection checklist — but the core edit workflow is a two-line summary with no pre-pack validation checkpoint, even though scripts/office/validate.py exists in the bundle yet is never referenced. Clear sequence with most checkpoints present and minor validation gaps fits anchor 4 rather than the fully-checkpointed anchor 5. | 4 / 5 |
Progressive Disclosure | Scored against the actual bundle structure: the Quick Reference table and section headers are well organized, but both top-level references — editing.md and pptxgenjs.md — point to files that do not exist in the bundle, so the one-level-deep disclosure structure is broken for two of the three task routes. Navigation signals are clear but lead nowhere, which is worse than the under-signaled references of anchor 3 and approaches the navigation failure of anchor 2. | 2 / 5 |
Total | 13 / 20 Passed |