taiwan-equity-research-coverage

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

• Third-party content exposure detected (high risk: 0.80). The SKILL.md explicitly states that Claude Code slash commands (e.g., /update-enrichment, /add-ticker, /discover) perform web searches ("3–5 web searches per ticker", "AI researches online") and yfinance usage fetches public finance data, meaning the agent will ingest open/public third‑party content (web pages/data) that can be used to generate enrichment applied to reports and thus materially influence actions.

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

• Potentially malicious external URL detected (high risk: 1.00). The skill's installation/runtime instructions explicitly fetch and run code from the external Git URL https://github.com/Timeverse/My-TW-Coverage (git clone ... then running python scripts), meaning remote code would be executed and controls runtime behavior.