terse-create

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

• Third-party content exposure detected (high risk: 0.75). The required runtime workflow reads outsider-authored free text in the “event context” path: webhook/sample trigger payloads (from external systems and potentially containing user-written content) are formatted via event.formatForAgentRunner() and inserted into generateText prompts (“Context: ${event.formatForAgentRunner()}”), enabling indirect prompt injection.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly instructs the agent at runtime to fetch https://docs.useterse.ai/llms.txt to discover and then pull specific docs pages, meaning remote content would be fetched and could directly control prompts/instructions.