paradox-sifter

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

• Third-party content exposure detected (high risk: 0.70). The skill’s workflow operates on a “literature corpus” and extracts “Limitations”/“Future Work” text from papers; if that corpus includes outsider-authored papers fetched/ingested at runtime (e.g., via web/API), that free-form prose would be read into the agent’s LLM context, creating indirect prompt-injection exposure.