crypto-research-desk

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

• Third-party content exposure detected (high risk: 0.75). The skill’s workflow explicitly requires a “NEWS / NARRATIVE” section and other research seats (e.g., price/trend, liquidity flows, macro) that typically fetch public web content or third-party data at runtime and then inject that readable text into the agent’s LLM context; the provided SKILL.md does not constrain those sources to user-authored material.