One Claude Code command audits your Managed Agents setup against Anthropic's security guide
brada · x · 2026-11-01
A practical tip for Claude Managed Agents: run /claude-api managed-agents-secure in Claude Code and it read-only compares your agent's settings with Anthropic's official security guide, returning MATCHES / DIFFERS / CAN'T TELL / DOES NOT APPLY per row plus the exact settings to change.
The guide's key recommendations, most important first:
- Run agents in cloud sandboxes unless tools need self-hosted infrastructure
- Set environment networking to limited with only required hosts
- Keep credentials in vaults; prefer Workload Identity Federation or workspace-scoped API keys
- Add alloweddomains allowlists to websearch and webfetch tools
- Use permissionpolicy: auto for read-only tools, alwaysask for irreversible actions
- Disable tools by default and enable only what's needed; agents reading untrusted input should never hold always-allowed write/execute tools
- Set budgets and a stop mechanism on each session and deployment
More from coding & agent
- Indie dev launches EasyComfy Pro: ComfyUI workflows as one-click presets with MCP support — Dependent-Heat3770 · 2026-11-01
- Human-in-the-Loop theater: Devs rubber-stamp Claude's 400-line 'dangerous' bash scripts — rickasaurus · 2026-11-01
- Claude-Code-Game-Studios: open-source template turns Claude Code into a 49-agent game studio — lxfater · 2026-11-01
- OpenMoat: open-source local policy checks to secure AI coding agents with machine access — __dotk__ · 2026-11-01
- Senior engineer: 'popularity is the weights of an LLM' — half of 600 engineering scars can't be baked into models — RandalSchwartz · 2026-11-01
- Factory's agent ships an MCP connector registry; david hoang wants Paper added — davidhoang · 2026-11-01