New Skills - August 31, 2026
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.
New Publisher Clusters - Guided This Sweep
Method Notes
- 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).
Evaluated and Queued
| Skill |
Source |
Installs |
Disposition |
| effective-html (6 skills) |
plannotator/effective-html |
10.7K |
Parked - below bar, no named publisher authority; hot momentum on html-wireframe; watch for growth |
| lark-apps, lark-vc, lark-openapi-explorer |
site/open.larksuite.com |
1+1 1H each |
Parked - site publisher (no GitHub repo segment), brand-watch queue |
| apifox-workflow-api-lifecycle |
site/apifox.com |
1+1 1H |
Parked - site publisher, brand-watch queue |
| byted-serverlessflink-volc-flink |
site/skills.volces.com |
1+1 1H |
Parked - site publisher, brand-watch queue |
Notable Signals for CorpusIQ
- 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.