Project
stig-ai-pipeline
Vendor-neutral tooling that turns DISA STIG XCCDF into structured, severity-ranked remediation (Markdown, JSON, and CSV), and can check a machine under a human approval gate. MIT licensed, free to run, no API key required.
At a glance
- MIT licensed; free; no API key required
- Two modules: a STIG parse/annotate pipeline, and stig-scan (human-approval gate; a safety gate refuses commands that change the system)
- Five published skills, including stig-to-tracker, stig-explain, and scan-related skills
- 84 tests
- Platform profiles for macOS, Linux, and Windows
- Seven official DISA STIGs parsed; 1,440 rules with plain-English explanations committed in-repo (stig-explain work dated 5–6 September 2026)
- Additional validation runs / tracker files dated 5 September 2026
- stig-scan work captured 6 September 2026 (commit 82b3057); includes a recorded scan of the author’s Mac against the macOS 26 STIG
- v0.4.0 adds a Windows extractor/checkpacks and a stig-scan-windows skill
Windows profiles and extractors are built and unit tested. A live Windows host scan has not been run.
Published validation (August 2026)
The HackerNoon article Building a Reusable AI Skill for STIG-to-Excel Conversion records validation across three system classes totaling 522 rules (Windows Server 2019, SQL Server 2022 Instance, Apple macOS 26) with no parsing anomalies.