Skills
Navigate
skills/avbel/ai-skills/codex-review-code/Gen Agent Trust Hub

codex-review-code

Pass

Audited by Gen Agent Trust Hub on Jul 18, 2026

Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill invokes the codex CLI and git via a bash script to gather repository state and perform reviews.
  • [SAFE]: The execution of the external codex tool is performed using --sandbox read-only, which is an OS-enforced boundary preventing the reviewer agent from writing to the filesystem or accessing the network.
  • [SAFE]: The skill implements robust protection against indirect prompt injection by using multi-layered fencing (long backtick sequences) for diffs and untracked file content, and includes explicit system-level instructions to treat that content strictly as data.
  • [SAFE]: Data handling is localized to the repository. The script captures the git diff and pipes it via stdin to the codex process, avoiding temporary files for sensitive diff data.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 18, 2026, 03:03 PM