AgentReadyHomeAgent Listing

← DeerFlow

DeerFlow — agentic threat model

8.1AIVSS 8.1 · High

DeerFlow presents a high agentic risk profile due to its multi-agent orchestration, long-horizon execution, and coding/tool capabilities, though its built-in sandboxed execution provides a critical layer of defense against host compromise.

OWASP AIVSS score rationale

AIVSS = (CVSS_Base + AARS) × Mitigation_Factor, where AARS = (10 − CVSS_Base) × (Factor_Sum / 10) × ThM
CVSS base 8.5AARS uplift 1.06Factor sum 7.1/10Threat ×1.0Mitigation ×0.85
Autonomy of Action
0.80
Goal-Driven Planning
0.90
Self-Modification
0.40
Dynamic Tool Use
0.80
Persistent Memory
0.80
Contextual Awareness
0.70
Dynamic Identity
0.30
Multi-Agent Interactions
0.90
Non-Determinism
0.80
Opacity & Reflexivity
0.70

Scored with the canonical OWASP AIVSS formula (AIVSS calculator reference); agentic risk factors estimated from the agent’s described capabilities.

MAESTRO 7-layer threat model

Per-layer threats for this agent. Layers tagged “not certain from listing” are general, caveated commentary where the public description didn’t pin that layer.

L1 · Foundation Models⚠ not certain from listing

Not certain from the listing — DeerFlow is a framework/harness and does not specify a default foundation model, though it likely supports various LLMs via API, exposing it to model-specific threats like prompt injection or adversarial alignment bypasses.

L2 · Data Operations⚠ not certain from listing

Not certain from the listing — While DeerFlow utilizes memory for long-horizon tasks, the specific vector database or data storage mechanism is not detailed, presenting risks of memory poisoning or data exfiltration.

L3 · Agent Frameworks✓ mapped

As a super agent harness orchestrating sub-agents, tools, and skills, it is highly susceptible to insecure tool integration, prompt injection leading to unauthorized tool execution, and memory poisoning during long-horizon tasks.

L4 · Deployment & Infrastructure✓ mapped

Specifically mentions sandboxed execution, which mitigates host compromise during coding/research automation, but vulnerabilities in the sandbox implementation or message gateway could allow container escape or lateral movement.

L5 · Evaluation & Observability⚠ not certain from listing

Not certain from the listing — The description does not detail evaluation, logging, or guardrail mechanisms, which could lead to blind spots in monitoring long-running agent tasks.

L6 · Security & Compliance (cross-cutting)⚠ not certain from listing

Not certain from the listing — No specific compliance certifications (like SOC2) or identity/access management policies are mentioned for this open-source framework.

L7 · Agent Ecosystem✓ mapped

Orchestrates multiple sub-agents via a message gateway, creating a high risk of agent-to-agent trust abuse, cascading failures, and rogue sub-agent behavior during complex, long-horizon tasks.

MAESTRO — the 7-layer agentic threat-modeling framework (Cloud Security Alliance / Ken Huang).