douyin-search-keyword

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 0.90). The code intentionally reads GUAIKEI_API_TOKEN (with a hardcoded fallback), and sends the token plus user search keywords and parameters to an external host (www.guaikei.com) via HTTPS query parameters while persisting results locally — this creates a high-risk credential/exfiltration and covert-collection pattern.

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

• Third-party content exposure detected (high risk: 0.85). High: the skill fetches outsider-authored public search results from the external service (network response body parsed from https.request in lib/douyin.js) and then formats that returned text fields (e.g., item.desc, author_nickname, etc.) into Markdown via douyin.formatMessage, which is printed and thus becomes LLM-readable context.

HIGH W008: Secret detected in skill content (API keys, tokens, passwords).

• Secret detected (high risk: 1.00). I scanned the repository for literal credential values. lib/key.js returns a 32-character hex string ("e10adc3949ba59abbe56e057f20f883e") as a fallback API token when GUAIKEI_API_TOKEN is missing/invalid. This is a concrete, literal token value (used as an API credential) rather than a placeholder or an obvious example password, and it will be sent to the upstream service. Therefore it meets the definition of a secret (hardcoded API token) and should be flagged. Other strings in the docs and examples are IDs/URLs or placeholders and are not flagged.