AgentReadyHomeAgent Listing

← guarded-whatsapp-mcp

guarded-whatsapp-mcp — agentic threat model

5.3AIVSS 5.3 · Medium

The guarded-whatsapp-mcp agent acts as a secure gateway for AI-driven WhatsApp communication, mitigating high-risk outbound messaging vectors through robust built-in controls like allowlisting and secret scanning. While its direct agentic autonomy is low, its role as a security proxy significantly reduces the threat of prompt-injection-led social engineering or data exfiltration.

OWASP AIVSS score rationale

AIVSS = (CVSS_Base + AARS) × Mitigation_Factor, where AARS = (10 − CVSS_Base) × (Factor_Sum / 10) × ThM
CVSS base 6.5AARS uplift 1.05Factor sum 3.0/10Threat ×1.0Mitigation ×0.7
Autonomy of Action
0.60
Goal-Driven Planning
0.20
Self-Modification
0.00
Dynamic Tool Use
0.40
Persistent Memory
0.20
Contextual Awareness
0.30
Dynamic Identity
0.50
Multi-Agent Interactions
0.40
Non-Determinism
0.20
Opacity & Reflexivity
0.20

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 — the agent is an MCP tool/server rather than a foundation model itself. It relies on upstream LLMs to generate message content, making it indirectly vulnerable to adversarial prompt injection that attempts to craft malicious messages.

L2 · Data Operations⚠ not certain from listing

Not certain from the listing — there is no mention of RAG, vector databases, or training data operations. It primarily processes transient outbound message payloads and recipient metadata.

L3 · Agent Frameworks✓ mapped

As an MCP tool, it integrates directly into agent orchestration frameworks. The primary threat is tool misuse or bypass if the host framework fails to strictly route all WhatsApp interactions through this guarded interface, or if prompt injection manipulates the agent into abusing allowed recipients.

L4 · Deployment & Infrastructure⚠ not certain from listing

Not certain from the listing — deployment security depends entirely on the host environment running the MCP server. However, the tool must securely handle and store sensitive WhatsApp API credentials and tokens, presenting a target for credential theft.

L5 · Evaluation & Observability✓ mapped

The agent strongly addresses this layer by incorporating built-in audit logging, rate limiting, and outbound secret scanning. These features mitigate observability blind spots and help detect anomalous or malicious outbound data flows in real-time.

L6 · Security & Compliance (cross-cutting)✓ mapped

Acts as a policy enforcement point (PEP) by implementing recipient allowlisting and data loss prevention (secret scanning). This provides a strong compliance and security control layer to govern agent communication boundaries.

L7 · Agent Ecosystem✓ mapped

Designed to govern messaging for AI agents within an ecosystem. It mitigates cascading risks where a compromised upstream agent might attempt to use WhatsApp to exfiltrate data or conduct unauthorized social engineering against external users.

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