image-remove-background

Pass

Audited by Gen Agent Trust Hub on Feb 19, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • EXTERNAL_DOWNLOADS (LOW): The local provider downloads a pre-trained model (Xenova/modnet) from Hugging Face (approx. 25MB) on first use. Hugging Face is a trusted organization, so this is considered a standard operational requirement.
  • CREDENTIALS_UNSAFE (SAFE): The skill requires API keys for cloud providers (FAL_API_KEY, REPLICATE_API_TOKEN) but correctly instructs users to use environment variables rather than hardcoding secrets.
  • PROMPT_INJECTION (LOW): The skill represents an indirect prompt injection surface as it ingests images from external URLs or local paths. Evidence Chain: 1. Ingestion points: --in parameter in SKILL.md. 2. Boundary markers: Absent. 3. Capability inventory: Image isolation and transformation. 4. Sanitization: Not specified in documentation.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 19, 2026, 01:20 AM