ai-content-pipeline

Fail

Audited by Gen Agent Trust Hub on Feb 25, 2026

Risk Level: HIGHEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [REMOTE_CODE_EXECUTION]: Fetches and executes the installation script for the infsh CLI from the vendor's official domain (cli.inference.sh).
  • [EXTERNAL_DOWNLOADS]: Downloads binaries and configuration files from the vendor's infrastructure (dist.inference.sh, cloud.inference.sh).
  • [PROMPT_INJECTION]: The skill design creates an indirect prompt injection surface by piping model outputs into subsequent inputs.
  • Ingestion points: Model outputs are saved and reused in the workflow.
  • Boundary markers: Absent in pipeline examples.
  • Capability inventory: Utilizes Bash(infsh *).
  • Sanitization: None detected in the provided patterns.
Recommendations
  • HIGH: Downloads and executes remote code from: https://cli.inference.sh - DO NOT USE without thorough review
Audit Metadata
Risk Level
HIGH
Analyzed
Feb 25, 2026, 05:35 PM