AGENT LAB: SKILLS

ai-image-generation

Fail

Audited by Gen Agent Trust Hub on Feb 19, 2026

Risk Level: CRITICALEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
  • Unverifiable Dependencies & Remote Code Execution (CRITICAL): The installation instructions in SKILL.md include curl -fsSL https://cli.inference.sh | sh. This pattern pipes a remote script from an untrusted domain directly into the shell, posing a critical security risk.
  • Unverifiable Dependencies & Remote Code Execution (HIGH): The skill recommends using npx skills add to install extensions from the untrusted inference-sh repository, which could lead to the execution of malicious code.
  • Indirect Prompt Injection (LOW): The skill lacks safeguards against malicious instructions embedded in user prompts. 1. Ingestion points: User input for image prompts. 2. Boundary markers: None present; input is interpolated directly into command lines. 3. Capability inventory: Broad Bash(infsh *) permissions allow network and tool interaction. 4. Sanitization: No input sanitization is performed.
Recommendations
  • HIGH: Downloads and executes remote code from: https://cli.inference.sh - DO NOT USE without thorough review
  • AI detected serious security threats
Audit Metadata
Risk Level
CRITICAL
Analyzed
Feb 19, 2026, 03:41 AM