devcontainer-setup
Creates and edits dev container configurations (.devcontainer/devcontainer.json) using trusted sources only.
Skill by ryan-minato
Install
Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.
npx agentmarket install skill/ryan-minato-devcontainer-setup
Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.
Get the weekly diff
The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.
About
Reviews
Write a reviewNo reviews yet. Installed it? Reviews come from verified installers only. How reviews work
Permissions
What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.
- Tools
- No tools found
- Secrets
- None asked for
- Install scripts
- None found
- Runs as
- Python package
Network hosts it talks to 1
codeload.github.com· found in install command
Files it touches
~/.claude/skills/devcontainer-setup
What's inside
- Skill name
devcontainer-setup- Description
- Creates and edits dev container configurations (.devcontainer/devcontainer.json) using trusted sources only. Use when creating or modifying a devcontainer — "set up a dev container", "make this repo open in a container", "configure Codespaces"; when a devcontainer.json is the file at hand; when a project needs a containerized or GPU/CUDA/ROCm development environment; or when asked what base images or features are available or why the dev container fails to build. Not for authoring or publishing Features, Templates, or prebuilt images — it consumes them.
- License
- Apache-2.0
- Compatibility
- scripts/list_sources.py requires uv and Python 3.11+ (stdlib only) plus network access to mcr.microsoft.com, ghcr.io, api.ngc.nvidia.com, and nvcr.io (--self-test checks these prerequisites). Verification is strongest with the devcontainer CLI and a container runtime; authoring alone needs neither.
- Instructions
- 146 lines in SKILL.md · view on GitHub
Before installing
- Ships executable scripts
Files (5)
- SKILL.md
- config-recipes.md · references/
- gpu.md · references/
- trust-domains.md · references/
- list_sources.py · scripts/
SKILL.md (first lines)
# Devcontainer Setup
Create or modify a project's dev container configuration
(`.devcontainer/devcontainer.json`) using trusted sources only. The policy
exists because images and features execute arbitrary code on the
developer's machine at build time — an unvetted source is a supply-chain
decision, not a style choice.
## Trusted sources policy
| Asset | Allowed origins |
|---|---|
| Base images | `mcr.microsoft.com/devcontainers/*`; NVIDIA NGC (`nvcr.io/nvidia/*`) |
| Features | `ghcr.io/devcontainers/features/*`; `ghcr.io/stacit-ai/devcontainer-features/*` |
| Templates | `ghcr.io/devcontainers/templates/*` |
These are the only sources to use unless the user explicitly requests a
specific other image, feature, or registry. An explicit user request
overrides the policy; never widen it on your own initiative (a convenient
community feature is not a reason), and when the user does override it,
note the deviation in your summary so it stays visible. When editing an
existing configuration that already references untrusted sources, leave
them in place but point them out.
## Assess the repository first
Before writing anything, check what already exists — the three
configuration axes (`image`, `build.dockerfile`, `dockerComposeFile`) are
peer options chosen by repository reality, not a default-to-image ladder:
- `.devcontainer/` already present → modify it; do not rebuild from
scratch.
- A `Dockerfile` exists → consider `build.dockerfile`, reusing or wrapping
it (its base image must still be trusted).
- A `docker-compose.yml` exists or development needs sidecar services
(database, queue) → consider `dockerComposeFile` attached to the
existing services.
- Otherwise → `image` + `features`, with the image matched to the
project's language/toolchain.
Templates are skipped by default; only for a brand-new, empty project
apply one from `ghcr.io/devcontainers/templates/*`. GPU/CUDA/ROCm work →
read [references/gpu.md](references/gpu.md). Dockerfile/compose recipes,
lifecycle hooks, mounts, ports, users, or `templates apply` → read
[references/config-recipes.md](references/config-recipes.md).
## Discover what the trusted sources offer
Enumerate the trusted sources with
[`scripts/list_sources.py`](scripts/list_sources.py) instead of guessing
names, options, or tags:
```bash
uv run scripts/list_sources.py --kind features --format json # all features + options
uv run scripts/list_sources.py --kind images --source ngc --filter pytorch
uv run scripts/list_sources.py --tags mcr.microsoft.com/devcontainers/python
```
The script only reaches trusted registries; a ref outside them exits 2.
`--format json` output is designed to be consumed directly. If the script
errors, or the user asks to add a trusted source, read
[references/trust-domains.md](references/trust-domains.md).
## Compose the configuration
Prefer `.devcontainer/devcontainer.json` (over a root `.devcontainer.json`)
and keep it minimal:
```jsonc
{
"name": "myproject",
"image": "mcr.microsoft.com/devcontainers/python:3-bookworm",
"features": {
"ghcr.io/devcontainers/features/github-cli:1": {},
"ghcr.io/stacit-ai/devcontainer-features/uv:1": {}
},
"postCreateCommand": "uv sync",
"customizations": {
"vscode": { "extensions": ["ms-python.python"] }
}
}
```
- Reference features by full OCI ref pinned to the major version shown by
the lister (`ghcr.io/devcontainers/features/python:1`); `{}` means
default options. Take option names from the lister's JSON output, never
from memory.
- Project dependency install belongs in `postCreateCommand`; every-start
services in `postStartCommand`.
- Pin the image tag from `--tags` output rather than trusting `latest` to
exist.
## Baseline features for non-prebuilt images
Only `mcr.microsoft.com/devcontainers/*` images (and images explicitly
prebuilt for dev containers with the needed features baked in) come with
the baseline a dev container expects. **Every other base image — NGC
images, plain language/distro images, the base under a custom Dockerfile —
must add:**
- `ghcr.io/devcontainers/features/common-utils:2` — non-root user, shell
and core utilities baseline;
- `ghcr.io/devcontainers/features/git:1` — and most projects also want
`ghcr.io/devcontainers/features/git-lfs:1`.
Skipping this breaks user creation, `remoteUser`, and other features in
ways that surface late. For GPU images the rule is mandatory in practice:
read [references/gpu.md](references/gpu.md) before writing any config
whose image comes from `nvcr.io` or that needs GPU access (NVIDIA and AMD
paths, `runArgs`, `hostRequirements.gpu`).
## Verify
1. Every image/feature/template ref in the config must resolve against the
lister's output (or be an explicitly user-approved exception).
2. If the `devcontainer` CLI is available, build it:
`npx -y @devcontainers/cli@0.87.0 build --workspace-folder .` (or
`devcontainer up` when a runtime is available) and fix what it reports.
3. If no CLI/runtime is available, say so explicitly and tell the user howFrom ryan-minato/skills, Apache-2.0.
More from ryan-minato/skills
- 0.570score
skill-authoring
ryan-minato
Skill lifecycle for this repository — layers repo-specific scaffolding, testing, and validation on top of generic skill authoring.
SkillAFreshness: AClaude Code +9 - 0.570score
change-workflow
ryan-minato
Runs this repository's change lifecycle on GitHub — the OpenSpec change loop, the draft pull request opened at the change record, approval on the draft…
SkillAFreshness: AClaude Code +9 - 0.565score
meta-harness
ryan-minato
Agent harness methodology — explains, designs, analyzes, audits, improves, and repairs everything that makes a project workable for agents.
SkillAFreshness: AClaude Code +9 - 0.560score
code-review
ryan-minato
Project-calibrated code review for this skill library — reviews a diff, PR number, branch, or path as instructions executed by agents, treats leaked secrets…
SkillAFreshness: AClaude Code +9 - 0.529score
ryan-minato-skills-installing
ryan-minato
Installs Agent Skills from the ryan-minato/skills library into a project or globally, and lists what the library offers.
SkillBFreshness: BClaude Code +9 - 0.529score
sensitivity-check
ryan-minato
Detects PII and leaked secrets or credentials in text or files, ending in a report with an explicit YES/NO conclusion.
SkillBFreshness: BClaude Code +9 - 0.529score
git-commit
ryan-minato
Executes the full git commit workflow as ordered safety gates before anything is committed.
SkillBFreshness: BClaude Code +9 - 0.525score
great-skill-writing
ryan-minato
Skill authoring — create, review, and repair Agent Skills.
SkillBFreshness: BClaude Code +9 - 0.525score
agentic-writing
ryan-minato
Agent-first documents — writes, reviews, improves, and prunes the files an agent reads before it works: AGENTS.md and other entrypoints, CLAUDE.md…
SkillBFreshness: BClaude Code +9 - 0.524score
plan-clarification
ryan-minato
Interrogates a plan, an idea, or a decision in rounds until the user and the agent hold one understanding of it, with nothing left silently assumed.
SkillBFreshness: BClaude Code +9 - 0.479score
human-writing
ryan-minato
Drafts, edits, and reviews text meant for human readers — articles, essays, announcements, documentation prose, newsletters — in English, Chinese, or Japanese.
SkillCFreshness: CClaude Code +9 - 0.479score
conventional-commits
ryan-minato
Drafts git commit messages in the Conventional Commits 1.0.0 format.
SkillCFreshness: CClaude Code +9
Signals
- Stars
- 0
- Forks
- 0
- Open issues
- 11
- Last commit (this folder)
- Sep 4, 2026 (1 month ago)
From GitHub (fetched Oct 24, 2026).
Details
- License
- Apache-2.0
- Repository
- github.com/ryan-minato/skills
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
- Source
- GitHubgithub-graphql
More from ryan-minato
- 0.570score
skill-authoring
ryan-minato
Skill lifecycle for this repository — layers repo-specific scaffolding, testing, and validation on top of generic skill authoring.
SkillAFreshness: AClaude Code +9 - 0.570score
change-workflow
ryan-minato
Runs this repository's change lifecycle on GitHub — the OpenSpec change loop, the draft pull request opened at the change record, approval on the draft…
SkillAFreshness: AClaude Code +9 - 0.565score
meta-harness
ryan-minato
Agent harness methodology — explains, designs, analyzes, audits, improves, and repairs everything that makes a project workable for agents.
SkillAFreshness: AClaude Code +9 - 0.560score
code-review
ryan-minato
Project-calibrated code review for this skill library — reviews a diff, PR number, branch, or path as instructions executed by agents, treats leaked secrets…
SkillAFreshness: AClaude Code +9 - 0.529score
ryan-minato-skills-installing
ryan-minato
Installs Agent Skills from the ryan-minato/skills library into a project or globally, and lists what the library offers.
SkillBFreshness: BClaude Code +9 - 0.529score
git-commit
ryan-minato
Executes the full git commit workflow as ordered safety gates before anything is committed.
SkillBFreshness: BClaude Code +9
Cite this page
devcontainer-setup skill. Agent Mart. https://agentmart.dev/skills/ryan-minato-devcontainer-setup (last updated 2026-10-24)