music-search

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The skill explicitly extracts and returns share passwords ("extractCode") in its JSON output, which requires the LLM to read and emit secret/password values verbatim, creating an exfiltration risk.

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

• Third-party content exposure detected (high risk: 0.85). Deep-search runtime path: music-search.js calls web-search to obtain candidate pageUrls, then callDeepExtract() passes those outsider-authored page URLs into deep_extract.py extract, which fetches arbitrary third-party HTML and extracts readable text (e.g., titles/nearby context for extractCode/format) into the JSON that is returned to the agent’s LLM context.

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

• Potentially malicious external URL detected (high risk: 0.85). The skill's runtime auto-install (ensurePythonEnv) creates a venv and runs pip to install requirements.txt (cloudscraper), which fetches and executes remote package code from PyPI (https://pypi.org/) as a required dependency for deep search.