byted-viking-search-knowledgebase

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

• Third-party content exposure detected (high risk: 0.85). Runtime path: scripts/viking_search.py calls /api/knowledge/collection/search_knowledge, then ingests outsider-authored knowledge-base chunk text from the API response fields like data.result_list[].content into the skill output (top_chunks), which is then available to the agent/LLM for answer generation (indirect prompt injection risk).