minimal-api-file-upload
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- Indirect Prompt Injection Surface: The skill provides instructions for building endpoints that ingest external data via file uploads. This represents a potential surface for indirect prompt injection or file-based attacks if an agent later processes the content of these files without sufficient isolation.
- Ingestion Points: Data enters the application context through
IFormFilein Step 1 andMultipartReaderin Step 5. - Capabilities: The instructions utilize file system capabilities, including
Directory.CreateDirectoryandFile.Createin Steps 4 and 5. - Sanitization Patterns: The skill proactively addresses common vulnerabilities by providing patterns for file signature (magic byte) validation, Content-Type verification, and safe filename generation using
Guid.NewGuid()to prevent path traversal in Step 4. - Boundary Markers: While the code includes validation logic to verify file integrity, it does not explicitly define prompt-level boundaries or "ignore embedded instructions" warnings for the AI agent itself, which is a standard consideration for agent-facing file processing.
Audit Metadata