competitor-research

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

• Third-party content exposure detected (high risk: 1.00). The skill explicitly fetches public app metadata from the iTunes Search API ("SERP top-200 per token" via https://itunes.apple.com/search) and the LLM is instructed to read and tokenize app titles/subtitles/keywords and to make a batched relevance judgement over each candidate's name + description[:200] (phase_a/phase_b JSON), so untrusted third-party content can directly influence keep/drop decisions and downstream actions.