oracle-codex
Pass
Audited by Gen Agent Trust Hub on May 10, 2026
Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes the
codexcommand-line utility through a dedicated wrapper script (scripts/run-codex-exec.sh). This wrapper enhances security by validating the model against an allowlist and hardcoding the sandbox mode toread-only.\n- [PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection as it processes external data like file contents and diffs.\n - Ingestion points: User queries and code context are interpolated into the prompt in
SKILL.md(Step 2).\n - Boundary markers: The instructions do not specify explicit delimiters for demarcating untrusted context from instructions.\n
- Capability inventory: The skill can execute complex code analysis tasks via the
codex execcommand.\n - Sanitization: There is no explicit sanitization of input data. However, the use of a single-quoted HEREDOC (
<<'EOF') in the execution script is a robust mitigation that prevents shell-level command injection from malicious inputs.
Audit Metadata