llm-streaming-response-handler
Pass
Audited by Gen Agent Trust Hub on Sep 18, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill references several well-known and trusted Node.js packages for LLM integration and cost estimation, such as
ai,@ai-sdk/openai, and@ai-sdk/anthropicfrom Vercel, andtiktokenfrom OpenAI.\n- [INDIRECT_PROMPT_INJECTION]: The skill provides patterns for processing real-time data from external LLM APIs, which inherently involves an indirect prompt injection surface as the agent processes potentially untrusted content.\n - Ingestion points: Data is received via the
fetchAPI stream reader in the provided code examples and thestream_tester.tsutility.\n - Boundary markers: The provided code snippets demonstrate standard streaming handling without explicit delimiter-based isolation of the untrusted model output.\n
- Capability inventory: The skill is focused on UI display and utility scripts; it does not grant capabilities for executing the streamed content as code or shell commands.\n
- Sanitization: Content sanitization is not explicitly implemented in the provided boilerplate patterns, leaving this responsibility to the developer using the skill.
Audit Metadata