The Security Skills Cluster Is Compounding

Leaderboard attention concentrates on the mega-collections at the top, but four security-work libraries have taken up mid-field positions — #44, #51, #53, and #63 on today's snapshot. Above them at #19 sits a different kind of repo — a scanner whose entire job is to vet skills like these before you install them.

The Big Picture

Star counts are today's leaderboard snapshot; the 30-day change compares August 4 and September 3 entries in SkillMap's tracking history.

Trend 1: Security-Work Skills Keep Compounding

The four libraries in the 2.5k-3.7k band cover different parts of the trade. gadievron/raptor configures Claude Code for offensive and defensive security work — rules, sub-agents, skills, and orchestrated security tooling aimed at adversarial thinking. snailsploit/claude-red is a curated library of offensive security SKILL.md files spanning SQLi through EDR evasion. elementalsouls/claude-osint packages external recon tradecraft — its description counts 8 skills, 100+ recon capabilities, and 80 secret-regex patterns for authorized red-team and bug-bounty engagements. ljagiello/ctf-skills covers the practice side: nine category skills for web exploitation, pwn, crypto, reverse engineering, forensics, and OSINT.

All four added stars over the last 30 days, and none of them is a newcomer — this is steady accumulation in a band the leaderboard's upper tier ignores.

Trend 2: Vetting Tooling Outgrew Its Own Category

nvidia/skillspector scans skills for vulnerabilities, malicious patterns, prompt injection, data exfiltration, and supply-chain risks before installation. SkillMap's history has it at 1,200 stars on June 5; today it stands at 15,746, #19 on the full leaderboard. The README cites research that 26.1% of skills contain vulnerabilities and 5.2% show likely malicious intent, and SkillSpector ships 71 detection patterns across 17 categories as part of NVIDIA's Verified Skills pipeline.

The interesting shape here: a security-scanning skill for the skills ecosystem itself is outgrowing most of the offensive libraries it would scan — and it is the natural companion to every repo in Trend 1.

Trend 3: Science Still Leads Raw Growth

Security is not the fastest story on the board. k-dense-ai/scientific-agent-skills added 9,621 stars between August 4 and September 3 (32,547 to 42,168), sitting at #7 overall — the strongest 30-day climb among the repos tracked in this article, and a reminder that domain-depth collections remain the ecosystem's biggest draw.

What SkillMap Published

Today's tutorial batch covers the hands-on side of this wave:

• ctf-skills: Nine Category Skills for Capture the Flag
• design-md-chrome: Turn Any Website's Design System Into a SKILL.md
• agent-skills-platform: Ship Skills Your Team Can Actually Trust
• Apify Agent Skills: Describe the Data, Let Your Agent Do the Scraping
• image-to-editable-ppt-skill: Rebuild Slide Images as Editable PowerPoint

Coming Next

Watch whether pre-install scanning becomes a default step rather than a best practice — if SkillSpector's trajectory holds, the next milestone is the leaderboard's top ten. And keep an eye on the 2.5k-3.7k band: it is where the last two months of accumulation have been quietest and steadiest.

Read the latest tutorials.