replicant

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

• Third-party content exposure detected (high risk: 0.75). Outsider free text can enter the LLM context when the workflow performs web/code search and then reads external repository source files from runtime-cloned repos (public web content / third-party repo text), which is then inspected and used as evidence.

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

• Potentially malicious external URL detected (high risk: 1.00). The skill explicitly runs git clone/fetch at runtime (e.g., git clone --depth 1 "https://github.com/$OWNER/$REPO.git" and variants like "git@github.com:owner/repo.git"), which pulls remote repository content that the agent then uses as source context to drive its prompts/answers.