A Category-by-Category Look at the Six Dimensions
Category averages hide more than they reveal — but the shape of what they hide is informative.
Averages across the 591 fully-evaluated skills, by category:
| Category | n | Total | Trigger | Structure | Workflow | Content | Engineering | Security |
|---|---|---|---|---|---|---|---|---|
| Documents | 18 | 8.86 | 8.42 | 9.01 | 8.79 | 9.11 | 8.64 | 9.33 |
| Video | 43 | 8.72 | 7.81 | 8.88 | 8.28 | 9.16 | 8.93 | 9.74 |
| Developer tools | 125 | 8.72 | 7.58 | 8.63 | 8.43 | 9.08 | 9.33 | 9.78 |
| Marketing | 119 | 8.58 | 9.30 | 7.68 | 8.97 | 8.44 | 5.61 | 9.60 |
| Writing | 36 | 8.53 | 7.56 | 8.54 | 8.12 | 8.78 | 9.18 | 9.54 |
| Productivity | 118 | 8.45 | 7.50 | 8.25 | 7.66 | 8.96 | 9.33 | 9.82 |
| Design | 38 | 8.43 | 7.38 | 8.43 | 7.81 | 8.82 | 9.24 | 9.64 |
| Data | 33 | 8.25 | 7.26 | 8.36 | 7.33 | 8.59 | 9.15 | 9.70 |
| Finance | 61 | 8.17 | 5.61 | 8.77 | 7.76 | 9.15 | 9.56 | 9.59 |
The two asymmetries worth looking at
Marketing is trigger-strong and engineering-weak. Trigger quality 9.30 — the highest of any category — against engineering 5.61, the lowest of any category by a wide margin.
That combination is coherent: a marketing skill is a written argument, and its description is the product. The packaging — runnable scripts, resolving paths, clean frontmatter — is secondary to the sales team that authored it.
Finance is the mirror. Trigger 5.61 (lowest), engineering 9.56 (near the top). Finance skills come from people who write code first and describe it second. The description is an afterthought; the working script is the point.
Both categories score in the middle overall (8.58 and 8.17) because the rubric weights (trigger 20, engineering 10) partially cancel the opposite strengths.
Documents scores highest on a small sample
Documents leads at 8.86, but n=18. The category is dominated by the official Anthropic skills — xlsx, docx, pptx — which are among the best-engineered files in the directory. That is an author effect, not a category effect.
The rule that keeps repeating
We made this point with nine categories and 591 skills, and it holds at every level of the data: category averages are author averages. Marketing's trigger score is not measuring marketing; it is measuring one large marketing catalogue written by people who are good at descriptions.
The practical consequence is that "which category is best" is the wrong question. The right question is "which skill, by whom", and the answer does not generalize upward.