webflux-test-reviewer

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

• Third-party content exposure detected (high risk: 1.00). The skill explicitly fetches and ingests public candidate repositories (user-provided GitHub URLs) — see scripts/setup_candidates.sh and the SKILL.md execution flow that instructs cloning/reading candidate directories and running git log — so the agent will read and act on untrusted, user-generated third-party content as part of its evaluation workflow.

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

• Potentially malicious external URL detected (high risk: 0.80). The repository includes scripts/setup_candidates.sh which runs git clone on arbitrary GitHub URLs (e.g., https://github.com/juseruhe/franchises) — at runtime this fetches remote code that is then read/injected into the agent's context to drive its evaluation prompts, so external content can directly control the agent's behavior.