gemini-review

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 1.00). The skill's workflows explicitly feed untrusted, user-generated pull request diffs into the model (e.g., "git diff origin/main...HEAD > diff.txt" and the GitHub Action/GitLab CI steps that call gemini with "$(cat diff.txt)"), so third-party PR/repo content is ingested and can directly influence the agent's review outputs and actions.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). Flagged because the skill explicitly runs the command "gemini extensions install https://github.com/gemini-cli-extensions/code-review" at runtime, which fetches and installs remote extension code that will execute locally and directly control the agent's review prompts/commands.