file-processing
Pass
Audited by Gen Agent Trust Hub on Oct 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to guide the processing of untrusted external files (such as CSVs, archives, and binary data), which creates a surface for indirect prompt injection if the content of these files influences agent behavior or includes malicious instructions.
- Ingestion points: The skill details ingestion through streaming CSV parsing (Step 1), direct S3 uploads via presigned URLs (Step 2), and general file processing.
- Boundary markers: The instructions recommend path sanitization and magic byte verification (Step 5 and Security Section) to delimit file types and structures, but does not specify prompts for the agent to ignore instructions embedded within the file data itself.
- Capability inventory: The skill facilitates file system I/O, S3 object management (GET/PUT/DELETE), and image manipulation using external libraries like
sharpandPillow. - Sanitization: The skill explicitly mandates several defensive layers, including asynchronous malware scanning (ClamAV), path sanitization for archive extraction, and strict magic byte checks to prevent type confusion.
Audit Metadata