upstash-box-py

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.90). The skill includes examples that embed API keys/tokens directly (e.g., "sk-...", "Authorization: Bearer ...", env={"API_KEY":"..."}) which would require an LLM to include secret values verbatim in generated code/headers — an exfiltration risk.

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

• Third-party content exposure detected (medium risk: 0.30). The SDK’s workflow invokes a first-party agent skill via skills=["upstash/qstash-js"] and runs it with untrusted, caller-supplied prompt/tool interactions (box.agent.run(...)), so outsider-authored free text can reach the agent skill path.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill configuration passes a GitHub repo as an agent skill ("upstash/qstash-js") which is fetched at Box.create time to extend/drive the agent, and it also allows attaching an external MCP server via the runtime URL "https://mcp.example.com/sse", both of which are runtime dependencies that can supply instructions or behavior to the agent.