prototype-design-html
Pass
Audited by Gen Agent Trust Hub on Jul 16, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill processes untrusted natural language input provided by users to generate executable HTML/JavaScript code. This creates a potential surface for indirect prompt injection, where malicious instructions embedded in a requirements document could attempt to influence the agent's behavior or the generated output.
- Ingestion points: User-provided requirements documents or natural language descriptions processed during the 'Analysis' phase (SKILL.md).
- Boundary markers: While no explicit delimiters (like XML tags) are mandated for the input text, the skill enforces a transformation process where requirements must first be converted into a 'Detailed natural language DSL' (references/prototype-dsl.md) before any code is generated.
- Capability inventory: The agent has the capability to write the generated HTML to the local filesystem (defaulting to output/prototype/html).
- Sanitization: The skill relies on the LLM's internal reasoning and the intermediate DSL translation phase to sanitize and structure the user's intent, reducing the risk of direct execution of malicious commands within the requirement text.
- [SAFE]: The skill instructions explicitly forbid the use of external dependencies such as CDNs, remote images, online fonts, or npm packages, ensuring that the generated output remains isolated and does not initiate unauthorized network connections.
Audit Metadata