code-reviewer

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.80). The skill explicitly instructs the agent to scan for "hardcoded secrets" in repository files and to include code examples in its suggestions while having read access (Read/Grep/Glob), which creates a realistic need to reproduce secret values verbatim in the review output.

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

• Third-party content exposure detected (medium risk: 0.30). The skill’s workflow reads outsider-authored PR inputs at “Context — Read PR description” and then proceeds to review the PR content, so the runtime ingests free text submitted with the pull request.