Promptus — agentic threat model

Uses the latest generative AI models for video, image, and 3D asset generation. Key threats include model reprogramming, adversarial inputs causing resource exhaustion, and intellectual property/copyright risks associated with generated outputs.

Not certain from the listing — No explicit details are provided regarding training data, vector databases, or RAG operations, though the system manages and stores generated digital assets, presenting risks of data exfiltration or asset lineage gaps.

Orchestrates generation via an intuitive ComfyUI node-based workflow builder ('Cosyflows'). This introduces significant risks of insecure tool integration, where malicious or poorly constructed custom nodes can execute arbitrary code during workflow runs.

Utilizes a decentralized GPU compute marketplace. This infrastructure model introduces severe threats of host compromise, lateral movement, and lack of sandboxing if workflows are executed on untrusted, external provider hardware.

Not certain from the listing — There is no mention of evaluation frameworks, output guardrails, or observability logging to detect anomalous workflow executions or malicious node behaviors.

Not certain from the listing — The directory does not specify identity management, access controls, or compliance standards (such as SOC2 or ISO) governing the marketplace or asset management system.

Operates within a decentralized marketplace ecosystem where users share workflows and compute resources. This creates exposure to supply-chain attacks (compromised workflows/nodes) and rogue compute providers manipulating execution outputs.