Post

Vulnify: Giving Your Agents a CVE Brain

Vulnify is an MCP-capable SQLite database that normalizes CVE data across authoritative sources (CVEProject, NVD, CISA KEV, EPSS, OSV) and exploit corpora (Nuclei, Exploit-DB, Metasploit) into a unified relational schema with 20 tables and foreign keys on CVE IDs as the spine. It solves the problem of non-deterministic web-based CVE lookups for pentesting agents by eliminating dependence on Tavily searches, scanner templates, or rate-limited vendor APIs. Agents can now answer deterministic questions like “Does this package version have this CVE?” and “Is there a public exploit?” with concrete sourced artifacts (PoC availability, Metasploit module status) without token waste or external calls. The article also documents the author’s spec-driven, style-enforced approach to building coherent agentic codebases: manual v0.1 to establish taste, agent augmentation to enforce patterns, then feature-slice development with written specs and stress-testing before implementation.

Read original article

This post is licensed under CC BY 4.0 by the author.