Security context
What an agent needs to avoid regressing past fixes and find the next vuln in this repo.
oumi-ai/oumi
main @ 5025b0f
2
Fixes
0
CVEs
HIGH
Peak severity
Highlights
Arbitrary Code Execution: 1 prior fix. Scrutinize any change in this area.
configs/skypilot/sky_chatqa.yaml: most-fixed (1 issue). Treat as high-risk during review.
1 high-severity fix in this history; regressions here are high-impact.
Recurring patterns
The bug types that recur here, drawn from past fixes, not open vulnerabilities.
Arbitrary Code Execution: Unsafe deserialization of PyTorch checkpoints using standard torch.load allows execution of arbitrary code via malicious pickle payloads embedded in model weights or optimizer states. Future model training and loading components must enforce safetensors or weights_only=True.
Credential Exposure: Hardcoding API keys or platform tokens (like Hugging Face write tokens) in yaml configuration files exposes administrative access to external platforms if codebases are publicized or shared.