Content
72%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.
A model of token efficiency and progressive disclosure: a lean routing body over real, well-organized reference files. Its weaknesses are execution-side — abstract phrasing, one garbled sentence, and no explicit validation/feedback loop for the whole-deck batch operations the skill itself mandates.
Suggestions
Add an explicit validation checkpoint and retry loop for whole-deck batch edits (e.g. after a batch change, verify every affected slide and re-run the batch if any slide fails acceptance).
Fix the garbled sentence "Initial/full authoring plans needs before layout" — it is currently unactionable.
Replace vague directives like "apply the shared media workflow, including live capabilities, authorized model selection, actual placement and plan/check outcomes" with pointers to the specific sections of shared-guidelines.md that define each step.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is ~24 lines with zero padding, no explanations of concepts Claude already knows, and every line carries a rule or routing instruction. It matches anchor 5 ("Lean and efficient; assumes Claude's competence; every token earns its place") — nothing can be trimmed without losing content. | 5 / 5 |
Actionability | There is real specificity in places — "stable `data-ipw-slide` roots", "semantic `--ipw-*` values", "separate `data-pptx-text`, `data-pptx-shape` and `data-pptx-image` objects", "`media/artifact_preview_review` is the sole client batch preview" — but much of the guidance is abstract ("apply the shared media workflow, including live capabilities, authorized model selection, actual placement and plan/check outcomes") and one sentence is garbled ("Initial/full authoring plans needs before layout"). That mixed profile fits anchor 3: some concrete guidance but incomplete, with key executable details deferred without precise signposting. | 3 / 5 |
Workflow Clarity | The "Read by task" section gives per-task routing, and partial validation is present ("check font fit", "distinguish checked, unavailable and unverified work"), but there is no explicit validate → fix → retry feedback loop, and the skill mandates whole-deck batch operations ("shared token/style changes require one whole-deck batch") with no stated validation checkpoint — per the scoring notes that caps workflow_clarity at 3. It is above anchor 2 because the sequence and per-task reading order are defined, but below anchor 4 because checkpoints are implicit rather than explicit. | 3 / 5 |
Progressive Disclosure | All three referenced files exist (references/slides-ppt.md, references/shared-guidelines.md, references/layout.md), references are one level deep (only sibling cross-links, no chains), and each is conditionally gated ("Read Layout selection only when selecting or recomposing a layout"), exactly matching anchor 5's clear overview with well-signaled one-level-deep references. slides-ppt.md is well-sectioned, and the body's routing maps cleanly onto those sections. | 5 / 5 |
Total | 16 / 20 Passed |