AI tool official security-tool
Risk level
low
10/100

Anthropic Sandbox Runtime ↗

Anthropic Sandbox Runtime is an officially published AI tool from Anthropic. Anomity scores its risk at 10 out of 100 (low).

Analysis summary

Anthropic Sandbox Runtime is an officially published AI tool from Anthropic. Anomity scores its risk at 10 out of 100 (low).

OS-level filesystem and network sandbox (srt) for AI agents, MCP servers and other processes, without a container

Why this entry scored 10 out of 100

This profile starts at the catalog's neutral base of 50. Because the entry is officially published, trust adjusts the score by -40. The entry declares no high-risk capabilities on its own, so capabilities add nothing. It matches no dangerous capability combinations. The clamped result is 10, placing it in the low band.

See the full scoring formula →

Findings

  • Published by the vendor Trust offset
    Contribution to score: -40

Configuration surface

Files this AI tool reads on the host. Anything written here changes what the agent does next.

  • srt-settings global json
    Sandbox settings: filesystem read/write rules, network domain allowlist, Unix sockets and isolation relaxations (.srt-settings.json)
    • win32${USERPROFILE}\.srt-settings.json
    • darwin${HOME}/.srt-settings.json
    • linux${HOME}/.srt-settings.json