ollama-batch-generation
Pass
Audited by Gen Agent Trust Hub on Oct 12, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill serves as a documentation and best-practices guide for batch processing with local models. It does not contain executable malicious scripts or risky automated behaviors.
- [INDIRECT_PROMPT_INJECTION]: The skill acknowledges the risk of malicious instructions within processed data (e.g., product names containing 'ignore the above'). It provides explicit defensive strategies:
- Ingestion points: Data from spreadsheets, web scrapes, and user input (specified in SKILL.md).
- Boundary markers: Instructions to explicitly label input as 'data, not instructions' within the prompt (specified in SKILL.md).
- Capability inventory: Local API calls to an Ollama server and local file system writes for caching and results (specified in SKILL.md).
- Sanitization: Defensive parsing using regular expressions to strip code fences and JSON raw decoding to ignore trailing text, followed by strict validation against schemas and banned-claims regex (specified in SKILL.md).
Audit Metadata