redis-semantic-cache

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

• Third-party content exposure detected (medium risk: 0.30). Redis LangCache cache-aside flow ingests outsider-authored free text as the user-provided prompt sent to POST /v1/caches/{cacheId}/entries/search, and the resulting cached response is returned without filtering beyond similarity/attributes.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill calls LangCache at runtime using the external endpoint (e.g., https://{os.getenv('HOST')}) to fetch cached LLM responses which are injected back as the agent's outputs, so the external URL directly controls the agent's responses.