The top 25 Claude Code skill authors by daily-eligible count
5,053 distinct authors have contributed at least one skill to the open ClaudSkills catalog. Only 531 have at least one skill that clears the daily_eligible quality threshold — the top ~12% of the catalog. And only 47 have ten or more. This is the leaderboard of those 47, ranked by absolute count of high-quality skills shipped. If you build polished, reusable Claude Code skills at volume, this is the list to land on.
The leaderboard
Note on Anthropic itself. The official Anthropic organization has 126 skills in the catalog, but only a handful currently clear the daily_eligible threshold. Anthropic's reference skills are intentionally minimal — they're seed examples, not productionized end-to-end playbooks. They don't appear on this leaderboard but they're browseable at /author/anthropic/.
Three patterns that compound
1. A consistent voice + a consistent SKILL.md template
Every author in the top 10 uses a recognizable template across their roster — same heading order, same anti-trigger section, same install snippet, same way of structuring the inputs/outputs block. The structural Quality Score rewards this directly (vocabulary diversity, length × structure, anti-trigger discipline). Once an author has internalized a template that the scorer rewards, every subsequent skill they publish gets a bump for free.
2. Verticals where the success criterion is unambiguous
The top authors lean heavily into integration, review, audit, and analysis. These are categories where "did it work?" is easy to answer (the connector either shipped or it didn't; the review either caught the bug or it didn't). Authors who work in unambiguous-success-criterion domains tend to write tighter SKILL.md files because they're closer to the underlying ground truth. Compare to skills in fuzzier domains like "creative writing assistance" — even excellent ones tend to score lower because the rubric necessarily reads differently.
3. Authoring at the team scale, not the individual scale
Several of the top 15 entries are organizations (Orchestra Research, OrchestKit, Crewm8, Vulnetix, FreedomIntelligence, ID8Labs, ClyptAI) rather than individuals. The organization-attributed entries hit the leaderboard because a team can sustain authoring volume an individual can't. If you maintain a skill library at work, attribute it consistently — the catalog will treat all your team's skills as one roster and the cumulative effect is visible.
How to land on this list
The leaderboard isn't gated. Anyone can break in. Practical advice:
- Pick one tight workflow you do every week and ship a SKILL.md for it. Don't write a 10k-word "do everything" skill. The scorer rewards focused skills with clear scope. Our SKILL.md authoring guide walks through the exact frontmatter fields the scorer reads.
- Always include a "When NOT to use this skill" section. This is the highest-leverage edit in the entire
structural_v2rubric. Skills with explicit out-of-scope statements consistently land in daily_eligible. - Publish under a consistent attribution. If you put your name in the
author:frontmatter field on three different skills, all three roll up into one author profile at/author/<your-handle>/. The catalog uses an alphanumeric-normalized display name as the bucket key, so "Jane Doe", "jane doe", and "Jane Doe <[email protected]>" all collapse into the same profile. - Submit the SKILL.md URL via the Submit form on the homepage. The miner picks up submissions on its next nightly run (01:00 local, typically finishes by ~05:00). If your skill passes the admission threshold (
TIER_A_MIN = 50on the structural heuristic) it appears in the catalog within ~24 hours. - Iterate on the same SKILL.md. The miner re-fetches public SKILL.md files when the upstream source signals a change. A second pass that tightens the language or adds the anti-trigger section usually moves the Pro Quality Score from "above admission" to "daily_eligible".
There are no fees, no acceleration tiers, no paid placement. The leaderboard ranking is a function of how many of your skills clear the same content-derived score that every other admitted skill is measured against.
Methodology
Author counts are computed from the live site/data/skills.json snapshot. The bucket key is an alphanumeric-normalized version of each skill's author frontmatter field, with the email suffix stripped. This collapses common variants (Jane Doe, jane doe, [email protected] → janedoe) into a single author profile, matching the /author/<handle>/ URL key used by the /author/ index.
The daily_eligible flag is set on the public payload for any skill whose Pro Quality Score is at least 80 on a 0-100 scale. Pro Quality Score blends 80% structural (anti-trigger discipline, length × structure, frontmatter completeness, vocabulary diversity, pricing/quota disclosure) with 20% metadata depth. Popularity signals (stars, forks, install counts) were dropped from the formula on 2026-05-07 because content-derived signals correlated more strongly with human quality judgment.
Ties broken by alphabetical author key. Re-rankings happen as the daily miner cycles repopulate the catalog — most authors' positions are stable cycle-to-cycle, but a single new high-quality batch can move the top 25 visibly.