AML.CS0054
Data Exfiltration via Remote Poisoned MCP Tool
- Type
- Exercise
- When
- 2025-04-01
- Actor, as ATLAS names it
- Not stated
- Target
- Model Context Protocol
- ATLAS attributes it to Snyk (security evaluation).
Techniques, in order
- 1LLM Prompt CraftingAI Attack Adaptation · AML.T0065
- 2Publish Poisoned AI Artifacts: AI Agent ToolsResource Development · AML.T0115.002
- 3Supply Chain Compromise: AI Agent ToolInitial Access · AML.T0010.005
- 4AI Agent Tool Poisoning: Definition and InstructionsPersistence · AML.T0110.000
- 5User Execution: Poisoned AI Agent ToolExecution · AML.T0011.002
- 6LLM Prompt Injection: IndirectExecution · AML.T0051.001
- 7AI Agent Tool InvocationExecution · AML.T0053
- 8Unsecured CredentialsCredential Access · AML.T0055
- 9AI Agent Tool Credential HarvestingCredential Access · AML.T0098
- 10Exfiltration via AI Agent Tool InvocationExfiltration · AML.T0086
- 11External Harms: User HarmImpact · AML.T0048.003
From MITRE ATLAS 2026.10 (Apache-2.0). ShieldMap lists who and what was involved and the techniques used, and links to ATLAS for the write-up; it does not copy the steps. Attribution is ATLAS's, not ShieldPi's.