executing-web-scraping

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.90). The skill tells the agent to "Copy the request as cURL" and to "reverse-engineer it with HTTPX" and also directs use of "requests/HTTPX + API key" or "Playwright with session cookies", which encourages taking Authorization headers, cookies, or API keys from captured requests and embedding them verbatim into generated code/commands, creating a high exfiltration risk.

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 1.00). This content provides explicit, actionable instructions for bypassing anti-bot and app protections (TLS fingerprint spoofing, Playwright stealth, proxy/CAPTCHA evasion), intercepting and reusing authentication material (cURL/cookies, storage_state), and SSL pinning bypass via Frida — techniques commonly used to achieve unauthorized access and credential theft.

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

• Third-party content exposure detected (high risk: 0.95). During the “LLM extraction” path (Step 3 router entry “IF unstructured… → LLM extraction”), the workflow fetches arbitrary target page HTML/JS-rendered content at runtime, converts it to Markdown, and then injects that outsider-authored page text into the LLM prompt as {markdown[:8000]} (references/llm-extraction.md).