Skip to content
Proposals/Grant sandbox access to ~/.docker, ~/.kube, and
proposaldeveloperP5Worth a lookGoogle Antigravity

Grant sandbox access to ~/.docker, ~/.kube, and ~/_netrc

Antigravity 1.4.6 blocks sandboxed reads/writes of ~/.docker, ~/.kube, and ~/_netrc unless you grant access, and adds in-place MCP variable setup in /plugin. Update those grants and finish any ⚠ needs setup plugins so Docker, Kubernetes, and MCP workflows keep working.

Why this loop

1.4.6 changes two workflows that break if you ignore them. Sandboxed commands can no longer read or modify ~/.docker, ~/.kube, or ~/_netrc unless you grant access; ~/.docker and ~/.kube used to be writable, so Docker registry and Kubernetes credential use inside the sandbox now needs an explicit grant. Installed plugins gain a Configure MCP servers... row in /plugin, so you can set or change MCP server variables without disabling and re-enabling; plugins with missing values show ⚠ needs setup, each server shows ✓ configured or the variables it still needs, and saving restarts enabled plugin servers. agy plugin validate now has --help, checks context providers (missing label or command, invalid or duplicate triggers, unknown fields, missing command files), warns about plugin.json skills entries that do not exist or contain no SKILL.md, and reports every SKILL.md the loader skips. /compact keeps attached images, PDFs, and other files in the summary. Allowing an action for the conversation now applies to the main agent and all subagents.

Proposed actions

  1. In Antigravity run /plugin, open every installed plugin marked ⚠ needs setup, choose the Configure MCP servers... row, fill every variable that is not marked ✓ configured, and save so enabled plugin servers restart with the new values.
  2. For sandboxed commands that use Docker registry or Kubernetes cluster credentials, explicitly grant sandbox access to ~/.docker and ~/.kube (and ~/_netrc if those credentials are used); 1.4.6 no longer allows sandboxed read or write of those paths by default.
  3. In each plugin repo run agy plugin validate --help then agy plugin validate, and fix missing context-provider label or command, invalid or duplicate triggers, unknown fields, missing command files, skills entries in plugin.json with no SKILL.md, and any SKILL.md the loader reports as skipped (via plugin.json or skills.json).
  4. After a long session that includes attached images or PDFs, run /compact instead of starting a new chat so those files stay in the conversation summary and remain visible to the agent.
  5. When a permission prompt appears in the main agent or a subagent, allow the action for the conversation once; that grant now applies across the main conversation and all of its subagents.

Agent prompt

Paste into your agent or query via MCP (get_agent_prompt) — free, no extra AI cost

Google Antigravity / agent task

DevAgentRadar → Google Antigravity

Goal: turn this release signal into a safe pilot plan for my stack.

Context

Assistant: Google Antigravity Proposal: Grant sandbox access to ~/.docker, ~/.kube, and ~/_netrc Summary: Antigravity 1.4.6 blocks sandboxed reads/writes of ~/.docker, ~/.kube, and ~/_netrc unless you grant access, and adds in-place MCP variable setup in /plugin. Update those grants and finish any ⚠ needs setup plugins so Docker, Kubernetes, and MCP workflows keep working. Primary source: https://github.com/google-antigravity/antigravity-cli/releases/tag/1.4.6

Why it matters

1.4.6 changes two workflows that break if you ignore them. Sandboxed commands can no longer read or modify ~/.docker, ~/.kube, or ~/_netrc unless you grant access; ~/.docker and ~/.kube used to be writable, so Docker registry and Kubernetes credential use inside the sandbox now needs an explicit grant. Installed plugins gain a Configure MCP servers... row in /plugin, so you can set or change MCP server variables without disabling and re-enabling; plugins with missing values show ⚠ needs setup, each server shows ✓ configured or the variables it still needs, and saving restarts enabled plugin servers. agy plugin validate now has --help, checks context providers (missing label or command, invalid or duplicate triggers, unknown fields, missing command files), warns about plugin.json skills entries that do not exist or contain no SKILL.md, and reports every SKILL.md the loader skips. /compact keeps attached images, PDFs, and other files in the summary. Allowing an action for the conversation now applies to the main agent and all subagents.

Suggested actions

  1. In Antigravity run /plugin, open every installed plugin marked ⚠ needs setup, choose the Configure MCP servers... row, fill every variable that is not marked ✓ configured, and save so enabled plugin servers restart with the new values.
  2. For sandboxed commands that use Docker registry or Kubernetes cluster credentials, explicitly grant sandbox access to ~/.docker and ~/.kube (and ~/_netrc if those credentials are used); 1.4.6 no longer allows sandboxed read or write of those paths by default.
  3. In each plugin repo run agy plugin validate --help then agy plugin validate, and fix missing context-provider label or command, invalid or duplicate triggers, unknown fields, missing command files, skills entries in plugin.json with no SKILL.md, and any SKILL.md the loader reports as skipped (via plugin.json or skills.json).
  4. After a long session that includes attached images or PDFs, run /compact instead of starting a new chat so those files stay in the conversation summary and remain visible to the agent.
  5. When a permission prompt appears in the main agent or a subagent, allow the action for the conversation once; that grant now applies across the main conversation and all of its subagents.

Config surfaces this release may change

  • MCP servers (high confidence) — check your repo before applying
  • skills (high confidence) — check your repo before applying
  • sandbox settings (high confidence) — check your repo before applying
  • context and compaction 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/google-antigravity-1-4-6-grant-sandbox-access-to-docker-kube-and-netrc and mark Applied, Skipped, or Failed. Proposal id: 3272d458-65bb-4a9a-b662-b2f5061b2005

Please:

  1. Map the change to concrete pilot steps
  2. Flag security / permission implications
  3. Keep the pilot reversible
agentmcpmodelsecuritycliRelease source ↗

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.

Originating release signal

Google Antigravity1.4.6Oct 23, 2026

1.4.6

Added a Configure MCP servers... row for installed plugins in /plugin, so you can set or change a plugin's MCP server variables at any time without disabling and re-enabling it; plugins with missing values show ⚠ needs setup, each server shows ✓ configured or the variables it still needs, and saving restarts an enabled plugin's servers so the new values take effect. · Improved terminal sandbox defaults to protect more credentials: sandboxed commands can no longer read or modify ~/.docker, ~/.kube, or ~/_netrc unless you grant access explicitly. · +10 more changes
Verified excerpt — the source's own words
  • Added a Configure MCP servers... row for installed plugins in /plugin, so you can set or change a plugin's MCP server variables at any time without disabling and re-enabling it; plugins with missing values show ⚠ needs setup, each server shows ✓ configured or the variables it still needs, and saving restarts an enabled plugin's servers so the new values take effect.
  • Improved terminal sandbox defaults to protect more credentials: sandboxed commands can no longer read or modify ~/.docker, ~/.kube, or ~/_netrc unless you grant access explicitly. Previously ~/.docker and ~/.kube were writable inside the sandbox, so sandboxed commands that use Docker registry or Kubernetes cluster credentials now need an explicit grant.
  • Improved context compaction, including /compact, to keep images, PDFs, and other files you attached in the conversation summary, so the agent can still see them after older history is summarized.
  • Improved permission prompts so that allowing an action for the conversation, in the main agent or in a subagent, applies across the main conversation and all of its subagents instead of asking again for the same action.

Excerpt ends here — this release continues at the source ↗.

Primary source ↗