Security tools (white hat)
Some skills are white-hat security tools. They're built for hacking tests, red-team work, capture-the-flag games and malware analysis. They do dangerous things on purpose: scan networks, break logins, run exploits. That's what they're for. Use them only for security work you're allowed to do.
These are not the same as the harmful collections. Those are skills built to trick an agent. White-hat tools are honest about what they do.
White-hat skills
Attack and testing
Forensics
Checked by hand
What the tag changes
- The verdict stays the same. A white-hat skill can still be use, caution or ignore. Of the tagged skills,
are use, caution and ignore (mostly copies). - The danger score stays the same.
of them score high risk, because they really do risky things. The tag explains why. - You can filter. Search has a "Security tools" filter: include, leave out, or only. Agents can pass
white_hat=include|exclude|only. Each result sayscategory: "white_hat_security".
The repos with the most white-hat skills. AeonDave/malskill, an offensive-security toolkit, has
What it gets wrong
- Some misses. A skill with no name or description, like a bare list of attack payloads, is often not tagged. In our hand check of risky skills,
of hacking notes got the tag. - Some wrong tags. In
random tagged skills, weren't security tools and were unclear. - Defensive work is left out on purpose. Skills that detect attacks, harden systems or check compliance are not tagged.
How we measured this
A skill is tagged when (a) its own name or description uses a clear offensive-security term (pentest, red team, CTF, privilege escalation, Metasploit, Burp Suite, Kerberoasting, malware analysis and so on) together with a second security word, or (b) it sits in a white-hat repo and its own text has a security word. White-hat repos: a reviewed list (AeonDave/malskill, grisuno/LazyOwn, dazeb/ai-supply-chain-lab and three SANS "Find Evil" forensics repos), plus white_hat.py; output out/full/white_hat.parquet and white_hat_stats.json.