gpt-model-effort-advisor

Warn

Audited by Gen Agent Trust Hub on Aug 8, 2026

Risk Level: MEDIUMCOMMAND_EXECUTION
Full Analysis
  • [COMMAND_EXECUTION]: The instructions in SKILL.md guide the agent to construct and execute shell commands using user-provided input, such as python3 registry.py query "<the user's task>". This pattern is vulnerable to shell command injection if a user provides a task description containing shell metacharacters like backticks (``), semicolons (;), or nested quotes.
  • [DATA_EXPOSURE]: The registry.py script performs local system inspection by reading the user's codex configuration file (~/.codex/config.toml) and walking the process tree using the ps command. While intended to detect the current session's model settings, this practice exposes local environment details and process command-line arguments to the agent's context.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Aug 8, 2026, 01:53 AM
Security Audit — agent-trust-hub — gpt-model-effort-advisor