Stats
Actions
Tags
Help us improve
Share bugs, ideas, or general feedback.
From prodsec-skills
Scan models for malicious code in model registries. Use when building, configuring, or reviewing model registry security, model ingestion pipelines, or model validation workflows.
npx claudepluginhub redhatproductsecurity/prodsec-skills --plugin prodsec-skillsHow this skill is triggered — by the user, by Claude, or both
Slash command
/prodsec-skills:model-registry-model-security-scanningThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Models stored in registries SHOULD be scanned to identify malicious code within them. This is a gate control applied at the registry level, before models are distributed to inference engines.
Scans AI models for malicious elements before loading in inference engines. Detects unsafe formats like pickle, backdoored models, and embedded scripts.
Detects compromised models from unverified sources, floating tags, and unreviewed registries with checksum verification, pinned revisions, and approval gates.
Detects AI/ML security vulnerabilities like unsafe model deserialization in PyTorch/Joblib/NumPy, prompt injection in LLM prompts, and risks in Jupyter notebooks or ML pipelines.
Share bugs, ideas, or general feedback.
Models stored in registries SHOULD be scanned to identify malicious code within them. This is a gate control applied at the registry level, before models are distributed to inference engines.
| Threat | Detection Approach |
|---|---|
| Pickle-based payloads | Detect Python pickle serialization that can execute arbitrary code |
| Embedded executables | Scan for binary executables or scripts hidden in model files |
| Unsafe serialization | Flag models using formats known to allow code execution on deserialization |
| Anomalous file structure | Detect unexpected files or metadata within the model package |
| Known malware signatures | Match against known malicious model signatures |
Model uploaded to registry
→ Signature verification (provenance check)
→ Security scanning (malicious content check)
→ If clean: mark as verified, make available
→ If suspicious: quarantine, alert security team
→ If malicious: reject, alert security team, log event