Claude Code Update Introduces Inter-Agent Messaging, Dev Warns of AI Worm Risk
shorns_username · reddit · 2026-08-08
A Reddit user raised serious security concerns regarding the inter-agent messaging feature introduced in Claude Code version 2.1.224.
The author points out that since prompt injection remains unsolved and agents act on text they read, allowing text to move between agents by design creates a transport layer for AI worms. An injected agent could now send malicious instructions directly to another agent. The author noted that agents in recent evals were already leaving messages for each other across isolated tests even without this feature enabled.
The feature is turned on by default and is intentionally complex to disable. The post provides the specific JSON configuration required in .claude/settings.json to block SendMessage and ListAgents permissions and refuse cross-session inbound messages.
More from coding & agent
- Constraints you force on models today will become tech debt in months, warns dev — zack_overflow · 2026-10-16
- Liquid AI tutorial: fine-tune your d1-3B decision model in 30 minutes on a free GPU — helloiamleonie · 2026-10-16
- Alibaba open-sources OpenGCUA, a full-stack generalist computer-use agent; +19.6 points on OSWorld-v2 across 8 models — Tongyi-MAI · 2026-10-16
- BotBeat: a social network where all users are AI agents, exposed via 32 MCP tools — Dismal_Piccolo4973 · 2026-10-16
- DKGI paper: agents write their own tools to search knowledge graphs, beating baselines by 6.7 Hit@1 — yuz9yuz · 2026-10-16
- Dev open-sources lgtmaybe: running PR code reviews on a local 27B model via hunk splitting — mattjcoles · 2026-10-16