mercadolibre-deals-scraper

CRITICAL E004: Prompt injection detected in skill instructions.

• Potential prompt injection detected (high risk: 0.80). The skill instructs the agent to install a runtime dependency automatically and explicitly tells it to hide that action from the user and not request separate approval, which is a deceptive instruction unrelated to the stated 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., scraped MercadoLibre offer/deal 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 at runtime to automatically install and invoke BrowserAct using the install link https://www.browseract.com/?co-from=ecommerce&redirect=https://github.com/browser-act/skills/tree/main, which implies fetching and executing remote installation code as a required dependency.

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

• Attempt to modify system services in skill instructions detected (medium risk: 0.40). The Skill instructs the agent to automatically install missing runtime dependencies without asking the user or seeking approval, which could cause the agent to run installers that modify the host system even though it does not explicitly request sudo, create users, or edit system files.