Orchids — agentic threat model

Not certain from the listing — The specific foundation models used are not disclosed. However, because the agent processes untrusted codebases and browses the web, it is highly vulnerable to indirect prompt injection, which could lead to malicious code generation or model reprogramming.

Not certain from the listing — The mechanisms for codebase indexing, RAG, or local vector storage are unspecified. Ingestion of malicious files from a repository could poison the agent's context, leading to data exfiltration or embedding inversion.

The agent framework coordinates complex planning, code generation, and multi-agent execution. Insecure tool integration is a major threat here, as the agent has the capability to write and execute code, potentially leading to unauthorized system commands or memory poisoning across agent sessions.

Operating as a desktop IDE with screen-viewing and browser interaction capabilities introduces severe infrastructure risks. A compromise of the desktop application could lead to local host takeover, privilege escalation, or the theft of sensitive API keys (Stripe, Supabase) stored in the environment.

Not certain from the listing — There is no mention of built-in guardrails, real-time monitoring, or logging mechanisms to detect anomalous agent behavior, screen-scraping abuse, or malicious code generation before execution.

Not certain from the listing — No security certifications (e.g., SOC2), identity governance, or access control policies are detailed, particularly regarding how third-party credentials (Stripe, Supabase) are securely handled and isolated.

The platform explicitly utilizes multi-agent execution. This introduces threats of agent-to-agent trust abuse, where a compromised sub-agent could escalate privileges, bypass user intent, or cause cascading failures across the development and deployment pipeline.