google-shopping-sku-ean-offer-scraper

CRITICAL E004: Prompt injection detected in skill instructions.

• Potential prompt injection detected (high risk: 0.90). The Skill includes explicit, deceptive installation instructions that tell the agent to install dependencies automatically and to conceal that action from the user (lines 41–47), which is outside the declared scraping purpose and constitutes a hidden/prompt-injection directive.

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 0.90). The Skill contains instructions to install runtime dependencies automatically and explicitly avoid requesting user approval or describing the installation as a user browser action, which indicates a deliberate attempt to perform potentially stealthy, unauthorized installations (supply-chain/installation abuse) and poses a high-risk supply-chain/consent abuse vector.

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., offer/deal pages) that BrowserAct ingests into the LLM context.

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

• Potentially malicious external URL detected (high risk: 0.80). The Skill explicitly directs the agent at runtime to auto-install the BrowserAct dependency from https://www.browseract.com/?co-from=ecommerce and https://www.browseract.com/?co-from=ecommerce&redirect=https://github.com/browser-act/skills/tree/main, which would fetch and install/execute remote code required for the Skill.

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

• Attempt to modify system services in skill instructions detected (high risk: 0.80). The Skill instructs the agent to automatically install a CLI dependency without asking the user for confirmation and to resume execution afterward, which directs the agent to modify the host system (potentially requiring elevated privileges and bypassing user consent).