Forty-six-query skills.sh API sweep (4,743 unique skills, zero query failures) with cross-reference against the full hermes/ tree, plus hot-leaderboard candidate discovery with publisher-page verification. 117 clusters known, 3 API candidates, 2 of which resolved to existing coverage (useosint/skills source-string variant of the guided osint-skills cluster; github/awesome-copilot already guided). The hot leaderboard surfaced a further 9 candidates; publisher-page fetches promoted 8 to drafts and parked the rest.
46-query API sweep returned 4,743 unique skills; the top-120 cluster diff flagged 3 candidates. useosint/skills (236.3K API sum) is the documented source-string variant of the already-guided useosint/osint-skills cluster - skipped. github/awesome-copilot (1.42M API sum) is already guided. dart-lang/skills was genuinely new (13.5K API sum undercounted the 144.0K publisher total 10x).
The hot leaderboard was the primary discovery channel this sweep, per the Aug 17 calibration: mohitmishra786/low-level-dev-skills (6 skills at +1 1H), basicmachines-co/basic-memory (3 skills), mengto/skills (2 skills), hubspot/agent-cli-skills, daymade/claude-code-skills, and lijigang/ljg-skills all surfaced via hot-page deltas the API never ranked.
Publisher-page totals are cited in guides (authoritative over API sums): dart-lang 144.0K vs 13.5K API sum, mengto 76.2K, daymade 58.4K, low-level-dev 41.5K, hubspot 17.8K, basic-memory 9.7K.
Below-20K drafting justifications: hubspot/agent-cli-skills on first-party hubspot org authority plus direct CRM domain relevance (github/gh-stack precedent); basicmachines-co/basic-memory on YC-backed vendor authority plus agent-memory domain relevance (makerskills suite-cohesion precedent).
Quality tiers follow audit badges: 🟢 where all three audits Pass (hallmark, dart-lang, mengto, basic-memory), 🟡 where Snyk (and Socket for hubspot) carry warnings, each named in the guide.
lijigang/ljg-skills ships Chinese-language SKILL.md content; the source language is named in the guide's Limitations per the non-English-suite precedent (wecomteam, genshijin).
Hallmark (nutlope, 27.6K⭐) encodes the anti-AI-slop design consensus CorpusIQ already enforces in content rules; it is the design-side counterpart to our content-voice gates.
HubSpot's official agent skills are the vendor's own CRM operations playbook - direct material for CorpusIQ's HubSpot connector and data-quality use cases.
Basic Machines' memory suite is the reference implementation of graph-first agent memory; its notes / observations / relations pattern maps onto how CorpusIQ agents structure session context.
AI answers on your data. See where every number comes from.
Connect your tools and ask in ChatGPT, Claude, or Perplexity. Enterprise-grade answers in minutes. From $29.95/month.