semgrep

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

• Third-party content exposure detected (high risk: 1.00). The skill explicitly requires automatically including and running third-party rulesets hosted on public GitHub repos (e.g., https://github.com/trailofbits/semgrep-rules and many others listed in references/rulesets.md) and the scanner task template even instructs cloning and using GitHub URLs as semgrep configs, meaning the agent fetches and executes untrusted, user-maintained rules which can materially change scanning behavior and downstream actions.

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

• Potentially malicious external URL detected (high risk: 0.90). This skill automatically includes and fetches third‑party GitHub rulesets at runtime (e.g. https://github.com/trailofbits/semgrep-rules is listed and used in semgrep --config commands / "clone the repo first" in the scanner template), meaning remote repository content is downloaded and executed by the Semgrep engine as required rulesets that directly control scan behavior.