Open source · MIT

Drillwright

Collect discipline-scoped cybersecurity training data and run inference against Foundation-Sec, Antares, and other local SLMs — one CLI from notes to fine-tune JSONL.

View on GitHub

What it does

Discipline packs

AppSec localisation, DFIR, detection engineering, cloud, threat intel, GRC, purple-team paths, AI security — each with schema and system prompt.

Collect & export

Interactive collect, folder ingest, and SFT / prompt-completion export ready for fine-tuning open security models.

Local inference

Talks OpenAI-compatible APIs — Ollama, vLLM, llama.cpp — so Antares and Foundation-Sec stay on your metal.

Install

pip install git+https://github.com/ibrahimmukherjee-boop/drillwright.git
drillwright packs
drillwright infer -d appsec_vuln_loc -p "SQLi in search endpoint…"

Built to sit beside Cisco’s open-weight security SLMs — not a fork of them. Your data, your endpoint, your fine-tunes.

Get Drillwright on GitHub