skills/dotnetage/mindx/open/Gen Agent Trust Hub

open

Pass

Audited by Gen Agent Trust Hub on Apr 11, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses the macOS open utility to launch files, applications, or URLs. It handles user-provided parameters securely by quoting variables and using jq for robust JSON parsing, which prevents shell command injection vulnerabilities.
  • [PROMPT_INJECTION]: The skill presents an attack surface for indirect prompt injection because it executes system actions based on potentially untrusted input (the target parameter).
  • Ingestion points: The target and app parameters in open_cli.sh receive input that could be controlled by an external source.
  • Boundary markers: The implementation lacks delimiters or warnings to prevent the agent from acting on instructions embedded within processed data.
  • Capability inventory: The skill can launch any local executable, open any file, or navigate to any URL via the open command.
  • Sanitization: The script does not validate the protocol of URLs or the location of files before passing them to the system utility.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 11, 2026, 10:48 AM
Security Audit — agent-trust-hub — open