ai-sdk-development
Pass
Audited by Gen Agent Trust Hub on May 2, 2026
Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [PROMPT_INJECTION]: The skill documents the implementation of AI agents that ingest untrusted data, which is an inherent surface for indirect prompt injection.\n
- Ingestion points: Data enters the agent context via user-supplied text passed to the
$agent->prompt()and$agent->stream()methods inSKILL.md.\n - Boundary markers: The provided code snippets do not demonstrate the use of delimiters or specific boundary instructions for the model prompts.\n
- Capability inventory: Agents can perform text generation, tool execution (including provider-level tools like
WebSearch), and manage files/vector stores as described in the SDK documentation.\n - Sanitization: The example logic does not include explicit input sanitization or validation routines before passing content to the AI providers.\n- [EXTERNAL_DOWNLOADS]: The skill mentions the installation of the
laravel/aipackage via Composer. This is the official, first-party SDK for the Laravel framework and aligns with the stated vendor identity.\n- [SAFE]: No malicious patterns, such as credential harvesting, unauthorized file access, or persistence mechanisms, were found. The skill serves its primary purpose as documentation for a legitimate development library.
Audit Metadata