transformers
Pass
Audited by Gen Agent Trust Hub on Mar 15, 2026
Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADSCREDENTIALS_UNSAFE
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill facilitates the installation of highly reputable and standard machine learning libraries such as
torch,transformers,datasets, andacceleratefrom official registries. - [CREDENTIALS_UNSAFE]: The skill outlines authentication for Hugging Face Hub using official methods and safe placeholders, following standard security practices for the service.
- [PROMPT_INJECTION]: The skill processes untrusted external data (text, images, and audio), creating a surface for indirect prompt injection.
- Ingestion points: Data processed via
pipeline,tokenizer, andTrainercomponents inSKILL.md. - Boundary markers: Absent; no specific markers or instructions are provided to treat processed data as untrusted or to ignore embedded instructions.
- Capability inventory: The skill includes the
TrainerAPI inSKILL.md, which possesses the capability to write results and model checkpoints to the local filesystem. - Sanitization: Absent; the skill does not implement validation or sanitization of the input data being processed.
Audit Metadata