inference-sh-cli

Fail

Audited by Gen Agent Trust Hub on Apr 4, 2026

Risk Level: HIGHREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The skill documentation recommends installing the CLI tool by fetching a script from an external domain and piping it directly to the shell (curl -fsSL https://cli.inference.sh | sh).
  • [COMMAND_EXECUTION]: The instructions direct the agent to interpolate user-provided prompts directly into shell command strings for the infsh tool. This pattern creates a significant risk of command injection if the input contains shell metacharacters.
  • [DATA_EXFILTRATION]: The skill includes functionality for uploading local files to a remote cloud service. This capability could be exploited to exfiltrate sensitive files from the environment if the agent is directed to do so by a malicious prompt.
  • [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it processes untrusted user data as input for shell-executing capabilities without using strict delimiters or programmatic sanitization.
  • Ingestion points: User-provided prompts and local file paths.
  • Boundary markers: None present.
  • Capability inventory: Shell command execution via the terminal tool and local filesystem read access.
  • Sanitization: Only a manual instruction to the agent to escape quotes is provided, which is insufficient for preventing injection attacks.
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
HIGH
Analyzed
Apr 4, 2026, 05:50 PM