link-reader

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The skill instructs the agent to prompt the user for ARK_API_KEY/MODEL_AGENT_API_KEY and write those secrets into a workspace environment file, which requires the agent to receive and output the secret values verbatim (high exfiltration risk).

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

• Third-party content exposure detected (high risk: 1.00). The skill's scripts and SKILL.md explicitly instruct running scripts/link_reader.py with arbitrary URLs and the link_reader function (which calls the AsyncArk LinkReader tool with the provided url_list) to fetch titles and content from web pages, PDFs, and Douyin—i.e., it ingests untrusted public third‑party content that the agent reads and could influence subsequent actions.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill fetches arbitrary external web pages at runtime (e.g., the example URL "https://example.com" passed to scripts/link_reader.py) and returns their content to the agent, which can directly inject remote text into prompts and thus control agent behavior.