Threat actor Storm-3358 uses Hermes agent framework with skills that train AI to evade SOC detection
cyb3rops · x · 2026-11-21
Security researcher fr0gger reports that threat actor Storm-3358 is using an AI agent framework called Hermes (alongside ARTEX) to compromise organizations.
Key detail: instruction files actively teach the agent to avoid triggering SOC alerts. A skill named redteam-detection-surface-and-resilience instructs: "use a read-only robot account, not admin, and throttle requests from '58 in 7 minutes' down to '10 in 30 minutes'." Screenshots show the skill file also grades actions by detection likelihood for red-team OPSEC purposes.
This shows attackers specifically engineering AI offensive frameworks to reduce their noise footprint. IOC: 199[.]19[.]95[.]68; files available in PromptIntel.
More from coding & agent
- Matt Pocock narrows his claim: up-front planning only for costly, large rework — mattpocockuk · 2026-11-21
- housecor: re-prompting from scratch now beats building it once the old way — housecor · 2026-11-21
- "Compiles" Isn't "Correct": Devs Push Back on a Common AI Coding Fallacy — gklambauer · 2026-11-21
- 55-85% of long Claude session usage was re-reading history — dev open-sources /flight-deck fix — PawelHuryn · 2026-11-21
- More or less up-front planning in the AI era? Two camps argue it out — housecor · 2026-11-21
- isometric-cities: coding agents turn real neighborhoods into pixel-art maps on Mac — zsakib_ · 2026-11-21