codex-agent-builder
Pass
Audited by Gen Agent Trust Hub on Apr 12, 2026
Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses local bash scripts (install_role.sh, write_role_config.sh, validate_role.sh) to automate the creation, installation, and validation of agent configuration files. These scripts utilize jq and yq for safe manipulation of JSON and TOML data.
- [DATA_EXFILTRATION]: No network exfiltration or sensitive data exposure patterns were detected. The skill instructions and templates include explicit guidance to redact secrets and avoid storing credentials in shared memory modules.
- [PROMPT_INJECTION]: Analysis of the skill instructions found no attempts to bypass safety filters or override agent behavior maliciously. The skill includes a dedicated evaluation case to test for and prevent prompt injection.
- [REMOTE_CODE_EXECUTION]: No remote code execution patterns, such as piped shell downloads or execution of untrusted remote scripts, were identified.
- [SAFE]: The skill implements robust validation logic for user inputs and configuration states. It employs secure data handling practices, such as using jq --arg for variable interpolation, which prevents command injection into the underlying shell scripts.
Audit Metadata