use-subagents-codex
Pass
Audited by Gen Agent Trust Hub on Sep 7, 2026
Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes a bundled Python script (
scripts/configure_worker_profile.py) to generate and install subagent profiles. The script implements robust input validation using regex for model identifiers and a restricted set of allowed reasoning efforts, preventing command injection or the creation of malicious configuration files. - [DATA_EXPOSURE]: The configuration utility manages files within the user's home directory (
~/.codex/agents/). It adheres to security best practices by explicitly setting file permissions to0o600(read/write access restricted to the owner) for both new configuration files and their backups. - [INDIRECT_PROMPT_INJECTION]: While the skill orchestrates multi-agent workflows which are susceptible to indirect injection, it includes defensive instructions for the parent agent to define self-contained briefs with explicit read/write scopes, boundary markers, and stop conditions for subagents.
Audit Metadata