serverless-deploy
Pass
Audited by Gen Agent Trust Hub on Oct 13, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted configuration data and metadata from external sources to automate infrastructure deployment and job execution.
- Ingestion points: The agent is instructed to read and classify content from HuggingFace repositories, ComfyUI workflow JSON files, and community templates found in the public catalog (
SKILL.md,reference/comfyui-serverless.md). - Boundary markers: There are no explicit instructions to use delimiters or ignore embedded natural language instructions when parsing external files such as
workflow.jsonor model metadata. - Capability inventory: The skill utilizes powerful capabilities including
create-endpoint,create-template,create-network-volume,run-endpoint(job execution), anddelete-endpoint(SKILL.md). - Sanitization: While the instructions describe correcting payload shapes based on error feedback, they do not specify sanitization or filtering of the text content within ingested files to prevent adversarial instructions from influencing the agent's logic.
Audit Metadata