Enable Kiro V3 local command sandbox with /sandbox
Kiro 2.31.0 gives every V3 user a local command sandbox and local autonomous mode. Contain agent shell commands (and optional local stdio MCP servers) in an OS sandbox, load MCP tools on demand, and keep V3 configuration and state under KIRO_HOME.Why this loop
2.31.0 makes command containment a default V3 capability, not a special setup. /sandbox puts agent-run shell commands and their child processes in an OS sandbox; you choose whether outbound network is blocked, limited to common development hosts, or open. Local stdio MCP servers can share that sandbox; the agent, built-in file tools, and remote MCP servers stay outside. Permission rules still apply, and the sandbox does not guarantee complete isolation, so proposed commands still need review. /autonomous on is for local V3 tasks where you want Kiro to proceed without per-command shell approvals: it switches to the Default agent and turns the sandbox on first. MCP tools now load on demand, so unused servers need not sit on the hot path. V3 configuration and state now live under KIRO_HOME—point sessions and docs there. Treat sandbox as the default for agent shells; use autonomous mode only on local tasks you already trust.
Proposed actions
- In a Kiro V3 session, run: /sandbox — then set outbound network to blocked or limited to common development hosts so agent shell commands and child processes run inside the OS sandbox.
- For a local V3 task that should proceed without per-command shell approvals, run: /autonomous on — Kiro switches to the Default agent and turns on the local command sandbox first.
- Start only the local stdio MCP servers you need, and run them inside the same /sandbox session. Leave remote MCP servers, the agent, and built-in file tools outside the sandbox; permission rules still govern those actions.
- Rely on on-demand MCP tool loading: do not preload unused MCP servers. Keep reviewing proposed shell commands even with the sandbox on, because isolation is not complete.
- Keep all V3 configuration and state under KIRO_HOME so 2.31.0 sessions read the same config and state layout.
Agent prompt
Paste into your agent or query via MCP (get_agent_prompt) — free, no extra AI costAGENTS.md / CLAUDE.md / GEMINI.md rule update
DevAgentRadar → Repo Harness Rule Patch
Goal: update our repository's permanent harness instructions based on this release.
Update the file that tool actually reads. Do not dump everything into one file.
- Claude Code → CLAUDE.md plus
.claude/(skills, hooks, settings, agents, commands). It does not read AGENTS.md natively; start CLAUDE.md with@AGENTS.md. - Codex, Copilot, Cursor, Factory Droid, Grok Build, Roo Code, Goose, OpenCode, Amp, Zed, Aider → AGENTS.md.
- Gemini CLI / Antigravity → GEMINI.md. AGENTS.md only if
context.fileNameis set. - Cursor glob-scoped rules →
.cursor/rules/*.mdc(plain.mdis ignored), not a second constitution. - Codex MCP →
.codex/config.toml, not.mcp.json. - Copilot extra instructions →
.github/copilot-instructions.md; custom agents →.github/agents/. - Windsurf →
.windsurf/rules(do not assume AGENTS.md). - Cline →
.clinerules. - Procedures → a skill (
SKILL.md). Enforcement the model must not skip → a hook. Isolated roles → subagents. - Do not fork the same rule into three tool files.
Context
Assistant: Kiro Proposal: Enable Kiro V3 local command sandbox with /sandbox Summary: Kiro 2.31.0 gives every V3 user a local command sandbox and local autonomous mode. Contain agent shell commands (and optional local stdio MCP servers) in an OS sandbox, load MCP tools on demand, and keep V3 configuration and state under KIRO_HOME. Primary source: https://kiro.dev/changelog/cli/2-31/
Why it matters
2.31.0 makes command containment a default V3 capability, not a special setup. /sandbox puts agent-run shell commands and their child processes in an OS sandbox; you choose whether outbound network is blocked, limited to common development hosts, or open. Local stdio MCP servers can share that sandbox; the agent, built-in file tools, and remote MCP servers stay outside. Permission rules still apply, and the sandbox does not guarantee complete isolation, so proposed commands still need review. /autonomous on is for local V3 tasks where you want Kiro to proceed without per-command shell approvals: it switches to the Default agent and turns the sandbox on first. MCP tools now load on demand, so unused servers need not sit on the hot path. V3 configuration and state now live under KIRO_HOME—point sessions and docs there. Treat sandbox as the default for agent shells; use autonomous mode only on local tasks you already trust.
Suggested actions
- In a Kiro V3 session, run: /sandbox — then set outbound network to blocked or limited to common development hosts so agent shell commands and child processes run inside the OS sandbox.
- For a local V3 task that should proceed without per-command shell approvals, run: /autonomous on — Kiro switches to the Default agent and turns on the local command sandbox first.
- Start only the local stdio MCP servers you need, and run them inside the same /sandbox session. Leave remote MCP servers, the agent, and built-in file tools outside the sandbox; permission rules still govern those actions.
- Rely on on-demand MCP tool loading: do not preload unused MCP servers. Keep reviewing proposed shell commands even with the sandbox on, because isolation is not complete.
- Keep all V3 configuration and state under KIRO_HOME so 2.31.0 sessions read the same config and state layout.
Config surfaces this release may change
- MCP servers — check your repo before applying
- sandbox settings — check your repo before applying
After you finish
Do not report this as applied to DevAgentRadar. You cannot write the visitor's loop.
Tell the human: open https://devagentradar.com/proposals/kiro-2-31-0-enable-kiro-v3-local-command-sandbox-with-sandbox and mark Applied, Skipped, or Failed. Proposal id: aaa3d75b-3446-4921-8801-72f029411458
Instructions:
- Read the existing context files and settings for the tools in this repo.
- Draft an explicit Git diff. Update ONLY agent harness rules or tool configuration.
- Call out deprecated flags, obsolete habits, or changed permission boundaries.
- Do not touch application logic.
Start with the proposed diff and say which layer and which tool's file it belongs in.
Your loop
This browser · no sign-in · not shared as “you”After you run the prompt
Only you can mark this. Agents cannot write your loop.
Your decision stays on this device. A public tally appears after a few votes.