Amoxt — agentic threat model

Not certain from the listing — likely utilizes commercial LLMs to generate project outlines and refine descriptions. Primary threats include prompt injection to manipulate cost/timeline estimates or bypass safety guardrails, and potential leakage of proprietary system instructions.

Not certain from the listing — allows users to upload files and context to improve descriptions. This introduces risks of data exfiltration of sensitive corporate documents, lack of isolation between tenant data, and potential processing of malicious payloads embedded in uploaded files.

Not certain from the listing — orchestration is limited to an interactive refinement loop. Threats include insecure state handling during the iterative prompt-refinement process and lack of input validation on user-defined automation requirements.

Not certain from the listing — hosted as a closed-source paid platform. Risks include standard web application vulnerabilities, insecure file storage for user uploads, and potential server-side request forgery (SSRF) if the platform attempts to parse external URLs provided in the context.

Not certain from the listing — no mention of active guardrails or observability tools. Gaps here could lead to undetected prompt injection attacks or failure to log unauthorized attempts to access other users' project scopes.

Not certain from the listing — closed-source vertical platform. Risks include compliance violations (e.g., GDPR/CCPA) if users upload PII within context files, and a lack of transparent access controls or audit trails for enterprise customers.

Not certain from the listing — operates as a standalone scoping assistant with human-in-the-loop development. The primary ecosystem threat is downstream supply chain risk, where compromised or manipulated project requirements lead to the creation of insecure automated workflows.