typescript-project
Pass
Audited by Gen Agent Trust Hub on Aug 6, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill references and downloads the Bun runtime installer from 'bun.sh' and the LiteLLM proxy from GitHub Container Registry ('ghcr.io/berriai/litellm'). These are official sources for the respective tools.
- [COMMAND_EXECUTION]: The setup instructions recommend installing the Bun runtime using a piped shell command ('curl | bash'). This is a well-known installation method for the Bun project.
- [EXTERNAL_DOWNLOADS]: The project templates specify the 'latest' version for core dependencies such as 'openai', 'zod', and 'typescript', which is presented as a deliberate versioning strategy for the architecture.
- [PROMPT_INJECTION]: The provided LLM adapter template in 'src/adapters/llm.adapter.ts' creates a surface for indirect prompt injection by passing user-supplied strings directly to the model API. Ingestion points: Untrusted data enters via 'prompt' and 'messages' arguments in 'complete', 'chat', and 'stream' functions. Boundary markers: The template does not include default delimiters or instructions for the model to ignore embedded commands. Capability inventory: The adapter provides capabilities for network operations to LLM providers. Sanitization: No validation or escaping of input data is implemented in the provided template code.
Audit Metadata