Samuele Giampieri is the creator, maintainer and AI platform architect of RedAmon, a self-hosted security-testing framework. The project joins reconnaissance, a structured map of findings and agent-assisted remediation in one inspectable workflow.[1]
Start with RedAmon →
Links
GitHub Contributions
Keep findings connected to the systems they describe
RedAmon chains reconnaissance tools and records discovered hosts, services, endpoints and relationships in a Neo4j graph. An agent can query that structured attack surface instead of receiving only disconnected scanner output. The documented pipeline continues through vulnerability triage to code changes proposed as GitHub pull requests. Its README distinguishes self-scored coverage and benchmark results from broader security assurance.[1]
Test instruction changes before keeping them
Giampieri’s Cortex work turns recurring corrections to a coding agent into replayable tasks. It proposes one skill or rule change at a time and compares repeated runs with and without that text. Changes are kept only when the measured gain passes the stated criteria. The paper’s tests also expose limits: unused text cannot establish whether harmful instructions were detected, a stronger model gained little in one setting, and a real project supplied too few corrections for learning.[4]
Tools, writing & talks
Tools
Start here
RedAmongithub.comA self-hosted reconnaissance, testing and remediation workflow with graph-backed findings and reviewable code changes.
Read the architecture and prerequisites before running it on systems you are authorized to test. Docker, substantial local resources and security-testing knowledge are needed; begin with an isolated target and inspect findings and proposed changes.
Creator and maintainer; copyright also names Ritesh Gohil[1][2]Original source
Writing, research & talks
Tests additions and removals to coding-agent instructions through repeated task comparisons, with explicit limits from the evaluation settings.
Author[4]