nebius-batch-synthetic
Fail
Audited by Socket on Mar 8, 2026
1 alert found:
Obfuscated FileObfuscated FileSKILL.md
HIGHObfuscated FileHIGH
SKILL.md
The Nebius Batch Inference skill appears coherent with its stated purpose. It uses standard, trusted tooling (pip-installed OpenAI SDK, Nebius API) and authenticates via an environment variable. Data flows follow a direct source-to-sink pattern (local prompts -> Nebius batch processing -> outputs -> training JSONL). No evidence of credential harvesting, unauthorized data exfiltration, or autonomous real-world actions. Overall, the footprint is proportionate and aligns with a legitimate bulk inference workflow; classified as BENIGN with low security risk.
Confidence: 98%
Audit Metadata