Hands-on playbook · 50 min
Model Scanner Playbook: Gating Model Files with ModelScan and picklescan, Exit Codes, Blind Spots and safetensors
A tested, hands-on guide to model file scanning: build safe test fixtures, read what ModelScan 0.8.8 and picklescan 1.0.5 really return, close the gaps we measured (silent skips, renamed pickles, unparsed files, a broken settings file), write a fail-closed CI gate, and move teams to safetensors.
- Build harmless test fixtures that scanners must flag, and see why they are safe
- Exit codes we measured: when 'No issues found!' still means the scan failed
- Four blind spots we reproduced, with the check that closes each one
- A fail-closed CI gate script, tested against 16 fixture folders
- Write a tiny pickle scanner in the standard library to see how they work
- Convert to safetensors without running the model's code on a laptop
- A 90-day rollout for model intake, plus an incident-style interview drill
- 01Under the hood
- 02Hands-on lab
- 03Recipes
- 04Rollout runbook
- 05Troubleshooting
- 06Interview drill
- 07Sources
Every worked answer in all 17 topics, every lesson in every course, the full concept curriculum, hands-on tool playbooks, the complete practice-test bank, and the downloadable security field guides.
6 months · One payment · No auto-renewal
An independent playbook for interview preparation and learning. AISecurityInterviews is not affiliated with or endorsed by Protect AI, now Palo Alto Networks (ModelScan); community (picklescan), and product names are trademarks of their owners. Use these procedures only on systems you own or are authorised to test, and check the vendor's documentation before changing production.