tensorslab-image
Pass
Audited by Gen Agent Trust Hub on Mar 2, 2026
Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection where malicious instructions embedded in user-provided content or external API responses could influence agent behavior.
- Ingestion points: User-supplied text descriptions processed in
SKILL.mdandreferences/scenarios.md, as well as image URLs returned by the TensorsLab API inscripts/tensorslab_image.py. - Boundary markers: Absent. The skill instructions for 'Agent processing' do not specify the use of delimiters or clear separation between user content and internal instructions.
- Capability inventory: The skill executes a Python script capable of making network requests (
requests.post,requests.get) and performing file system writes to the user's home directory (~/tensorslab_output). - Sanitization: No evidence of sanitization or escaping of external content before it is processed or passed to the API client script.
- [EXTERNAL_DOWNLOADS]: The script
scripts/tensorslab_image.pydownloads generated image files from the vendor's API endpoint (test.tensorai.tensorslab.com) to the local directory~/tensorslab_output.
Audit Metadata