claude-remote-peers
Pass
Audited by Gen Agent Trust Hub on Oct 17, 2026
Risk Level: SAFECOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill relies on shell commands including
ssh,mkdir, andrmto set up and tear down the messaging bridge. These commands are executed via thebunshell andBun.spawnas part of the tool's core functionality. - [REMOTE_CODE_EXECUTION]: A hardcoded Python shim is transmitted to and executed on the remote host via SSH stdin. This script is used to discover active Claude Code sessions by reading the remote registry and checking process statuses.
- [INDIRECT_PROMPT_INJECTION]: The tool processes session metadata (JSON) from the remote machine. It implements safety checks by verifying that the paths provided by the remote registry do not conflict with existing local files, which prevents the bridge from being used to overwrite sensitive system files or SSH keys.
Audit Metadata