Skills
skills/frankxai/creator-skills/acos-visual-gen/Gen Agent Trust Hub

acos-visual-gen

Pass

Audited by Gen Agent Trust Hub on Jul 4, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface (Category 8c). It retrieves factual data via a WebSearch tool and interpolates that data directly into a prompt for the mcp__nanobanana__generate_image tool without explicit sanitization or strict boundary markers. This could allow maliciously crafted content on an external website to influence the generated image's visual content.
  • Ingestion points: WebSearch output is used in Step 1 to populate facts.
  • Boundary markers: Absent. Facts are interpolated directly into the {fact_n_visualized} placeholders.
  • Capability inventory: Image generation via mcp__nanobanana__generate_image tool.
  • Sanitization: Not specified in the instructions.
  • [SAFE]: No obfuscation, data exfiltration patterns, or unauthorized command executions were detected. The skill uses standard MCP tool invocation patterns.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 4, 2026, 10:56 PM
Security Audit — agent-trust-hub — acos-visual-gen