laravel-ai-sdk

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.80). Yes — SKILL.md's mandatory Agent Workflow instructs spawning an agent (mcp__context7__query-docs) to "Pull latest laravel.com/docs/13.x/ai-sdk" (a public website), and the docs also describe FileSearch and provider-managed file/vector stores (OpenAI/Anthropic Files API) whose retrieved content is fed back to the model as part of tool calls, meaning public/untrusted third‑party content is ingested and can materially influence agent decisions.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.80). The workflow mandates spawning an agent (mcp__context7__query-docs) to pull the latest examples from https://laravel.com/docs/13.x/ai-sdk at runtime, which means external content from that URL would be fetched and used to shape agent prompts/instructions.