productreview-au-scraper

CRITICAL E004: Prompt injection detected in skill instructions.

• Potential prompt injection detected (high risk: 1.00). The Skill explicitly instructs the agent to install a dependency automatically and to conceal that action (do not ask the user to confirm or describe it as a user browser request), which is a deceptive instruction outside the declared scraping purpose.

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

• Third-party content exposure detected (high risk: 0.85). SKILL.md instructs the agent to use the BrowserAct CLI to “access real browser pages” and “capture browser-visible page content” at runtime, which can include outsider-authored free text from public web pages (e.g., arbitrary product/review pages) that the LLM ingests.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly instructs the agent to auto-install a required runtime dependency using https://www.browseract.com/?co-from=ecommerce&redirect=https://github.com/browser-act/skills/tree/main, which would fetch and install external code required for operation.

MEDIUM W013: Attempt to modify system services in skill instructions.

• Attempt to modify system services in skill instructions detected (medium risk: 0.60). The Skill explicitly directs the agent to automatically install the BrowserAct CLI as a required runtime dependency and to proceed without asking the user for confirmation, which instructs the agent to modify the host system by performing unattended software installation (a state-changing operation), even though it doesn't explicitly request privilege escalation or edits to protected system files.