ai-content-pipeline

Fail

Audited by Gen Agent Trust Hub on Feb 21, 2026

Risk Level: CRITICALREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • REMOTE_CODE_EXECUTION (CRITICAL): The SKILL.md file includes a command to install the CLI using curl -fsSL https://cli.inference.sh | sh. Piped execution of remote scripts from untrusted domains is a critical security vulnerability as it allows for arbitrary code execution with the user's privileges.\n- EXTERNAL_DOWNLOADS (HIGH): The skill initiates downloads from https://cli.inference.sh and recommends adding more skills via npx skills add inferencesh/skills. This domain is not recognized as a trusted source, presenting a risk of software supply chain compromise.\n- COMMAND_EXECUTION (MEDIUM): The skill requires the Bash tool and performs multiple infsh command executions to process data and generate content, creating a significant attack surface.\n- PROMPT_INJECTION (LOW): The skill implements a multi-step pipeline where the output of one model is fed directly into another tool, making it vulnerable to indirect prompt injection.\n
  • Ingestion points: Processes data from multiple intermediate files like script.json, summary.json, and various image/video JSON outputs.\n
  • Boundary markers: No delimiters or safety instructions are used when interpolating model outputs into subsequent command inputs.\n
  • Capability inventory: Uses Bash to execute platform tools and manage file outputs.\n
  • Sanitization: There is no evidence of sanitization or validation of the AI-generated content before it is processed in the next stage of the pipeline.
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 21, 2026, 04:56 PM