AGENT DEVELOPMENT / Docker
Docker Agent run
Pins Docker's skill for docker agent run, including the safety modes and the exception that a local stdio MCP server stays outside the sandbox VM.
“Use this skill when running a Docker Agent with `docker agent run`, choosing a safety/approval mode, using the `--sandbox` isolation flag, setting up aliases, or troubleshooting a run (missing credentials, worktrees). Even if the user just says they want to "run my agent", "make my agent auto-approve tool calls", "run this agent safely", or "why can't my agent see my API key", this skill applies. Covers `--safety` (strict/balanced/restricted/autonomous), `--yolo`, `--sandbox` and its network allowlist, `--worktree`, `docker agent alias`, and `docker agent doctor`.”
01 / THE REASONING
Why this made the selection.
- The catalogue has Docker's agent-config and sandbox-credential skills and not this run skill.
- The file says unattended runs should use restricted rather than autonomous, and it states that sandboxes are reused across runs.
02 / THE REVIEW RECORD
What we actually inspected.
Source review has boundaries.
A clear record is more useful than a “safe” badge.
Material inspected
- skills/docker-agent-run/SKILL.md
- README.md
- LICENSE
Our findings
- The skill defines strict, balanced, restricted, and autonomous safety modes, and it says autonomous is equivalent to --yolo.
- It says not to default an unattended run to autonomous or --yolo, and to use restricted so an unexpected tool call fails closed.
- It says --sandbox keeps built-in toolsets in a microVM, with the exception that a local stdio MCP server still runs as a host process, and that sandboxes are reused rather than torn down at the end of a session.
Not established by this review
- The resource was not executed and its tools were not called.
- docker agent run was not invoked.
- No sandbox was started.
- Not executed: it was not installed or run.
The review applies to the material and revision named here. A newer upstream release can change its behavior.
03 / PUT IT TO WORK
Add a skill to your workflow.
DOCUMENTED COMMAND
npx skills add docker/skills Run in your terminal to choose your AI tools, project or global scope, and installation method. This short command follows the current upstream version; the review covers the revision shown above.
- The README's quick start is the command above and says it prompts for which skills and agents.
- A non-interactive example on the next lines names docker-compose-patterns, not docker-agent-run.
- The product table describes docker-agent-run as the skill for safety modes, sandbox isolation, aliases, worktrees, and troubleshooting.
Before you start
- A skills-compatible agent
- The docker-agent CLI plugin if the commands are run, which the skill says needs Docker Desktop 4.63+ or a standalone install
- The sbx CLI if --sandbox is used
04 / FOLLOW THE EVIDENCE
The source trail.
Our notes are separate from the original resource.
Check upstream before adopting a new version.
- docker-agent-run SKILL.md ↗Checked
https://github.com/docker/skills/blob/f79172733725e2fd236343ee7b963c446ed85d09/skills/docker-agent-run/SKILL.md
Supports: safety modes and sandbox MCP exception
- Repository README ↗Checked
https://github.com/docker/skills/blob/f79172733725e2fd236343ee7b963c446ed85d09/README.md
Supports: install line and docker-agent-run table text
- Repository LICENSE ↗Checked
https://github.com/docker/skills/blob/f79172733725e2fd236343ee7b963c446ed85d09/LICENSE
Supports: Apache License Version 2.0
https://raw.githubusercontent.com/docker/skills/f79172733725e2fd236343ee7b963c446ed85d09/skills/docker-agent-run/SKILL.md
Supports: upstreamDescription, review.notTested, review.reviewedAt
https://raw.githubusercontent.com/docker/skills/f79172733725e2fd236343ee7b963c446ed85d09/README.md
Supports: install.command