adaption-ai
Pass
Audited by Gen Agent Trust Hub on Jun 28, 2026
Risk Level: SAFE
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill requires the 'adaption' Python package, which is the official SDK for the Adaption AI service.
- [COMMAND_EXECUTION]: Provided scripts like 'e2e_pipeline.py' and 'async_pipelines.py' automate dataset workflows including file uploads, job monitoring, and result retrieval.
- [DATA_EXFILTRATION]: The skill transmits dataset files to 'api.adaptionlabs.ai' for augmentation and evaluation. This transfer is necessary for the core functionality as described in the documentation.
- [CREDENTIALS_UNSAFE]: Authentication is managed via environment variables ('ADAPTION_API_KEY'), and code examples correctly use a non-functional 'pt_live_...' placeholder.
- [PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection as it ingests data from local files, Hugging Face, and Kaggle. 1. Ingestion points: 'e2e_pipeline.py' (file argument) and SDK import methods. 2. Boundary markers: Not implemented in the provided Python examples. 3. Capability inventory: SDK methods for file upload, status polling, and data download. 4. Sanitization: No client-side sanitization is performed as data is processed by the external service. This surface is considered safe given the intended use case of dataset adaptation.
Audit Metadata