ai-content-pipeline

Fail

Audited by Gen Agent Trust Hub on Feb 18, 2026

Risk Level: CRITICALREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • REMOTE_CODE_EXECUTION (CRITICAL): The skill instructs the user to execute curl -fsSL https://cli.inference.sh | sh. This is a classic remote code execution (RCE) vector where a shell script is downloaded and immediately executed without prior review.
  • Evidence: curl -fsSL https://cli.inference.sh | sh in SKILL.md.
  • EXTERNAL_DOWNLOADS (CRITICAL): The domain inference.sh is not a trusted source according to security protocols. Downloading and running executables from unknown third-party domains poses a high risk of malware infection.
  • COMMAND_EXECUTION (HIGH): The skill makes extensive use of the Bash tool to run the infsh CLI. This provides a broad attack surface where the agent could be manipulated into executing unintended system commands.
  • PROMPT_INJECTION (LOW): The skill is vulnerable to Indirect Prompt Injection (Category 8).
  • Ingestion points: Processes external data such as <blog-content> and <script-text> in the 'Blog to Video Pipeline'.
  • Boundary markers: Absent. External data is interpolated directly into JSON payloads for LLM tools.
  • Capability inventory: Access to network-enabled tools (infsh app run) and file system operations (redirecting output to .json files).
  • Sanitization: None detected. The skill assumes input data is benign.
  • DYNAMIC EXECUTION (MEDIUM): The skill uses shell loops and variable interpolation to build command strings at runtime (e.g., infsh app run ... --input "{\"prompt\": \"Visual representing point $i: <point-text>\"}"). This can lead to command injection if the variables contain shell metacharacters.
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 18, 2026, 05:45 AM