cninfo-to-notebooklm

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

• Third-party content exposure detected (high risk: 0.95). Runtime path: scripts/download.py downloads outsider-authored PDF reports from public cninfo.com.cn (http://static.cninfo.com.cn/{adjunct_url}) and then scripts/upload.py uploads those PDFs as sources into NotebookLM, making their extracted text available to the agent’s LLM context.

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

• Secret detected (high risk: 1.00). I scanned the repository files for literal, high-entropy credentials.

Findings:

• In scripts/download.py the class CnInfoDownloader hardcodes cookies:
• "JSESSIONID": "9A110350B0056BE0C4FDD8A627EF2868"
• This is a 32-character hex-like string that looks like a real session identifier (high entropy) and could grant access if valid. It meets the definition of a secret and should be treated as sensitive.
• "insert_cookie": "37836164"
• This is a short numeric value (low entropy). Per the rules (ignore simple/low-entropy values and setup-like values), I am not flagging it as a secret.

No other high-entropy API keys, private keys, tokens, or passwords were found in the skill content. Header values, URLs, CLI examples, prompts, and README examples are documentation and not secrets.