UTC

AML.CS0051

OpenClaw Command & Control via Prompt Injection

Type
Exercise
When
2026-02-03
Actor, as ATLAS names it
Not stated
Target
OpenClaw
  • ATLAS attributes it to HiddenLayer (security evaluation).
Techniques, in order
  1. 1Search Open Websites/Domains: Code RepositoriesReconnaissance · AML.T0095.000
  2. 2Acquire Public AI Artifacts: AI Agent ConfigurationResource Development · AML.T0002.002
  3. 3Discover LLM System Information: Special Character SetsDiscovery · AML.T0069.000
  4. 4Discover LLM System Information: System Instruction KeywordsDiscovery · AML.T0069.001
  5. 5LLM Prompt CraftingAI Attack Adaptation · AML.T0065
  6. 6Acquire InfrastructureResource Development · AML.T0008
  7. 7Stage CapabilitiesResource Development · AML.T0079
  8. 8MasqueradingDefense Evasion · AML.T0074
  9. 9Drive-by CompromiseInitial Access · AML.T0078
  10. 10LLM Prompt Injection: IndirectExecution · AML.T0051.001
  11. 11LLM JailbreakDefense Evasion · AML.T0054
  12. 12AI Agent Tool InvocationExecution · AML.T0053
  13. 13Modify AI Agent ConfigurationPersistence · AML.T0081
  14. 14LLM Prompt Injection: DirectExecution · AML.T0051.000
  15. 15AI Agent Context Poisoning: ThreadPersistence · AML.T0080.001
  16. 16AI AgentCommand and Control · AML.T0108
  17. 17Machine Compromise: Local AI AgentImpact · AML.T0112.000
The full case study on MITRE ATLAS →

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.