Skip to main content

Project · Working demo

GRC News Assistant

a cybersecurity news feed, tiered by an LLM for GRC relevance, that I built and run myself.

What it is

A pipeline that fetches a curated set of cybersecurity sources, uses an LLM to tier each article by GRC relevance on a five-tier scale, and publishes the result as a public live feed on this site. The output reads like something an IT auditor would actually keep open in a browser tab, rather than a firehose of everything cybersecurity.

Why

Most GRC portfolio pieces stop at frameworks and policy templates. I wanted to apply the same judgement to a system that actually runs, with its own code, containers and security controls. Thinking like that is easier to show on something live than to describe, and harder to argue with.

How it works, briefly

The workflow runs on n8n, in an isolated container on my own machine. It processes sources one at a time, so a single misbehaving feed cannot poison the batch. Each article passes through an LLM-assisted tiering step, and the results are emitted as a public JSON feed that the live-feed page on this site consumes. Container hardening includes a read-only filesystem, a non-root user, dropped Linux capabilities, and encrypted credential storage. The specific sources, the tiering prompt, and version details are held back on purpose.

Credit

Adopts a five-tier scoring framework in the spirit of Steve McMichael's Simply Cyber GRC News Assistant work, tuned for financial services IT audit context. Steve's original build inspired the project shape; the implementation, sources, prompt, and hardening choices here are my own.

Full risk assessment (NIST CSF 2.0 structured, following NIST SP 800-30 methodology) · coming as part of v1.5