codex-computer-use

Pass

Audited by Gen Agent Trust Hub on Jul 15, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes the OpenAI codex CLI tool (codex exec) to perform verification tasks involving browser and simulator interaction.- [COMMAND_EXECUTION]: Uses the tool's sandbox flag -s danger-full-access to grant the executor necessary permissions for desktop automation and GUI access.- [PROMPT_INJECTION]: The skill establishes an orchestration pattern where instructions are passed to a sub-agent, creating an indirect prompt injection surface.
  • Ingestion points: Self-contained prompts generated for codex and passed via a temporary file.
  • Boundary markers: The instructions do not define specific delimiters to separate user-influenced data from task instructions.
  • Capability inventory: The executor tool possesses capabilities for desktop interaction, browser automation, and simulator control.
  • Sanitization: The workflow does not implement explicit sanitization or validation of the prompt content passed to the CLI.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 15, 2026, 03:21 AM
Security Audit — agent-trust-hub — codex-computer-use