supermemory

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

• Third-party content exposure detected (high risk: 0.90). This skill explicitly ingests arbitrary public URLs/web pages and other third‑party content (see references/api-reference.md POST /v3/documents and quickstart.md examples like adding "https://example.com/documentation") and then has the agent read and inject those retrieved memories/search results into system prompts and decision logic (see SKILL.md/quickstart and SDK integration examples), so untrusted web content can materially influence agent actions.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill performs runtime calls to the Supermemory API (e.g., https://api.supermemory.ai
• /v3/documents, /v4/search, /v4/memories, profile()) and directly injects the returned profile/memories into system prompts, so external content from https://api.supermemory.ai can directly control agent prompts and is a required dependency.