sap-ai-core

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

• Third-party content exposure detected (high risk: 1.00). The skill explicitly ingests external, potentially untrusted content — e.g., grounding pipelines that pull from SharePoint, S3, SFTP (references/grounding-rag.md) and GitHub app syncs (references/advanced-features.md) — and then injects retrieved chunks into templating prompts ({{$context}}) that are sent to LLMs, so third‑party content can materially influence agent behavior.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill exposes an Applications/Git Sync feature that routinely fetches workflow and declarative prompt templates from a GitHub repository at runtime (e.g., https://github.com/org/ai-workflows), and those synced templates directly control prompts/workflows the platform runs, making this an active runtime dependency that can modify agent instructions.