html-api-sdk
Pass
Audited by Gen Agent Trust Hub on Jun 27, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The documentation describes usage patterns for reading data from local files and passing it directly into Large Language Model (LLM) prompts, establishing a surface for indirect prompt injection attacks. \n- Ingestion points: The window.Magic.fs.readFile API (detailed in SKILL.md and demonstrated in references/complete-examples.md) allows micro-apps to load content from workspace files. \n- Boundary markers: The provided code examples do not demonstrate the use of delimiters (such as XML tags or triple quotes) or specific instructions to prevent the agent from obeying commands embedded within the data. \n- Capability inventory: The SDK provides extensive capabilities including file system writes (fs.writeFile), updating agent input messages (setInputMessage), and creating new communication topics (project.createTopicAndSend). \n- Sanitization: There is no documentation or examples provided regarding the validation or sanitization of file content before it is interpolated into LLM prompts.
Audit Metadata