Super-Authors — The Handful of Names Behind a Third of Our Top 50
Evaluate enough skills and you start seeing the same names. We finally counted: the top 50 slots in our directory belong to fewer than twenty authors. Three of them hold twenty-nine of the fifty.
Our directory holds 637 skills from hundreds of authors. The median author ships one or two. But sort by score and look at the author column of the top 50, and the distribution collapses:
- calesthio — 11 skills in the top 50 (including the entire HyperFrames family)
- JimLiu — 10 skills (the baoyu toolkit)
- Jeffallan — 8 skills (the Muse memory system and related)
- anthropics — 6 (the official Office trio and API references)
- everyone else: one or two
Three authors, twenty-nine of the top fifty. This isn't the category-average effect — it's its sharper cousin: the score distribution is dominated by a handful of people who write skills as a craft, repeatedly.
What the three share
They write systems, not files. Calesthio's HyperFrames family is a ten-skill pipeline with a router and a contract. JimLiu's baoyu toolkit is a twenty-one-skill bilingual content pipeline. Jeffallan's Muse system is a memory architecture — capture, organize, recall, connect — where each skill is a component of one model of how memory should work. Strip the branding and each author has a thesis about how agents should work, expressed across files.
They specialize hard. Video rendering. Bilingual publishing. Memory. Zero overlap between the three portfolios — and near-zero overlap within each. The interference problem that kills average multi-skill authors (five skills describing the same thing) doesn't exist here, because each author knows their domain well enough to know where one job ends and the next begins.
Their floor is high. Calesthio's family runs 5.7–9.7 (one deliberate outlier); JimLiu's runs 8.9–9.5; Jeffallan's Muse set runs 9.0+. The interesting number isn't the ceiling — it's the floor. A 9.7 with a 5.7 sibling is one great skill; a family that never drops below 8.9 is a process.
What separates them from the 1.4-skill author
Nothing about resources — skill files are just markdown. The separation is in how they treat the trigger field. The weakest dimension across our whole directory is trigger quality (7.74 average), and it's exactly where super-authors spend their effort: each skill's "use when" is specific enough to win the loading decision against its own siblings. Average authors write triggers that describe the skill; super-authors write triggers that discriminate.
There's also a volume-skill interaction most authors get backwards. Shipping twenty skills doesn't make you good at skills — shipping twenty skills that don't overlap forces the trigger discipline that makes all twenty good. The portfolio shape drives the quality, not the other way around.
The anthropics counterexample
The most instructive row in the list is anthropics — six skills, all official, all 9.5+. Their advantage isn't secret knowledge; the Office trio (docx 9.7, xlsx 9.8, pptx 9.7) scores on the same rubric as everyone else. What they do differently is scope: each official skill covers exactly one format, states its boundaries explicitly, and refuses to generalize. The 9.8s are the most modest skills in the directory.
Why this matters for readers
Practical browsing shortcut: author is a quality signal, category isn't. When you find one skill you like, check the author's other entries before searching the category — the odds of adjacent quality are dramatically higher. Every skill page in our directory links the author's portfolio.
And for authors: the path to a 9.5 isn't writing one brilliant skill. It's picking a domain you know deeply, defining its job boundaries, and writing every skill inside those boundaries until the family is complete. The score follows the system.