batch-api-job-processing
Pass
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill provides a standard implementation for batch processing using the OpenAI Python library. The use of local file operations for JSONL creation and polling for batch status represents common and legitimate development practices.
- [INDIRECT_PROMPT_INJECTION]: The
create_batch_jobfunction processes external data (provided in theitemslist) and interpolates it directly into an LLM prompt. This creates a potential surface for indirect prompt injection if the source data contains malicious instructions, though this is a standard design pattern for the skill's intended purpose of bulk data transformation. - Ingestion points: The
itemsparameter in thecreate_batch_jobfunction withinSKILL.md. - Boundary markers: Absent; the data is directly interpolated into the message object.
- Capability inventory: Local file writing and network communication with OpenAI's API.
- Sanitization: None; the skill implementation passes data directly to the API provider.
Audit Metadata