New Skills - September 10, 2026 (Morning)¶
Forty-three-query skills.sh API sweep (3,775 unique skills; 2 queries failed - social media and content - and were compensated with targeted re-fetches that surfaced this batch), cluster-level cross-reference against the full hermes/ tree, hot-leaderboard coverage check (clean, 185 ranked / 32 sources, all catalog-covered), and the tiered cross-reference sweep (606 unique / 65 NEW / 99 PARTIAL, PARTIAL ≥100 empty, zero query errors). One new publisher cluster cleared the drafting bar.
New Skills¶
| Skill | Publisher | Installs | Category | What It Does |
|---|---|---|---|---|
3 skills (x-scraper, instagram-scraper, tiktok-scraper) |
apidojo-io/social-media-skills | 8,026 combined | Social Media / Data Extraction | Agent skills for scraping X (Twitter), Instagram, and TikTok data at scale via Apify actors - tweets by query/profile/hashtag/thread/date range, Instagram posts/comments/followers, TikTok videos/profiles/music. Standard Agent Skills format, npx skills add install |
Setup Guides¶
Evaluated and Skipped¶
| Cluster | Installs | Reason |
|---|---|---|
mathews-tom/armory (prompt-lab) |
77 | Claude Code / Claude.ai package collection - adapters generated for Cursor, OpenAI Codex, and Gemini CLI; zero Hermes targeting; no HERMES.md. Claude-family rejection class |
vm0-ai/vm0-skills (arga-labs) |
9 | Agent Skills format, but install docs target Claude Code Marketplace / GitHub Copilot; no Hermes mention; far below the 40-install floor. Below-bar park (watch) |
| its-a-skill-issue/superpowers | 9,589 combined (ai-video-generation 8,055 / agent-browser 1,534) |
Republished inference.sh mirror #5 - README is verbatim inference.sh skill repo text (evening-pass verification). Canonical coverage: halt-catch-fire/skills → inference-sh-skills-setup.md. Mirror class: 101-skills, magentosh, qu-skills, skills-shell, its-a-skill-issue |
awesome-skills/agent-skills-manager (agent-skills-manager) |
5 | Chinese symlink-based multi-agent skills manager (dotey scheme; Claude Code / Codex / Pi / OpenCode / Hermes via symlink + rsync). Hermes named on the badge, but 5 installs is far below the 40-install floor. Below-bar park (watch) |
All 63 other NEW flags from the tiered cross-reference mapped to the standing-rejection roster recorded in the corpusiq-docs-management skill run-log; no other bar-clearers, no overturns.
Evening pass (Sep 10, 20:00 MST): 44-query sweep (4,004 unique skills, zero failed queries - collector DEFAULT_QUERIES fix applied: social media + content split into social / media / content marketing), cluster diff 119 known / 1 candidate, tiered cross-reference 607 unique / 63 NEW / 100 PARTIAL (PARTIAL ≥100 empty), hot board clean (183 ranked / 33 sources, all catalog-covered). No bar-clearers; the two evening-pass records above were added to this table.
Security Notes¶
The one drafted cluster is 🟡 tier: the publisher has no Trust Hub / Socket / Snyk audit verdicts published on skills.sh - stated honestly in the guide with a review-first recommendation. The SKILL.md files are short, single-purpose API-pattern documents (Apify actor sync/async curl patterns) and manually reviewable in minutes.
Source¶
skills.sh API sweep (43 queries, 3,775 unique skills + compensated re-fetch of the 2 failed queries - the same social media / content pair that failed on Sep 9 and hid the Salesforce / Claude SEO mega-clusters), hot leaderboard scrape (clean), cluster-level cross-reference against hermes/, tiered cross-reference sweep (15 queries). Full playbook: corpusiq-docs-management skill, references/daily-skills-sweep-cron.md + references/sweep-run-log.md.