Why a blog?
A lot of what we build lives in READMEs and skill files: methodologies, attack patterns, guardrail configurations. That format is great for machines and for people already deep in a project, but it leaves no room for the why—the design decisions, the trade-offs, the dead ends. This blog is that missing space.
What to expect
Expect articles roughly along these lines:
- Offensive security — how Agent-Smith chains skills across recon, exploitation, and reporting, and what it means to teach an LLM methodology instead of scripts.
- Defensive security — building semantic allow-lists with Seraph, LLM-as-judge patterns, and defending against prompts nobody has seen yet.
- Methodology — mapping work to OWASP ASVS 5.0, the LLM Top 10, MITRE ATT&CK, PASTA, and STRIDE, and turning frameworks into practical, chainable skills.
- Field notes — findings, write-ups, and lessons from real assessments (sanitized, of course).
Join in
Everything we do is open source and community-driven. If an article sparks an idea, disagreement, or a pull request, come find us on Discord or the GitHub org.
See you in the next post.