setup-claude-devcontainer

Files (6)

• SKILL.md
• install-block.sh · templates/
• network-manifest-entries.json · templates/
• post-start-block.sh · templates/
• readme-bullet.md · templates/
• yolo-alias-block.sh · templates/

SKILL.md (first lines)

# Setup Claude Code Devcontainer

Adds Claude Code to a devcontainer already generated by the
[setup-devcontainer](../setup-devcontainer/SKILL.md) skill. This skill never
touches `devcontainer.json` or the `Dockerfile` — it appends its own
marker-keyed content to `post-create.sh`, `post-start.sh` (if skill sync is
opted into), and `README.md`, and idempotently ensures its own Network
Manifest entry exists in `.devcontainer/network-manifest.json`.

## 0. Security & Trust

This skill's generated automation touches three areas an external audit may flag. Each is
opt-in, scoped to this container, and exists for a specific reason — none of it runs on the
host, only inside the generated devcontainer.

- **Installer Provenance**: installs Claude Code via Anthropic's own official installer
  (`https://claude.ai/install.sh`, documented at code.claude.com/docs/en/quickstart). Runs
  once, at container creation (`postCreateCommand`), not on every start — and only if
  `~/.local/bin/claude` doesn't already exist.
- **Agent Authority**: the optional `claude-yolo` alias uses `--permission-mode auto
  --worktree --remote-control`, deliberately not `--dangerously-skip-permissions` — that flag
  triggers Claude Code's own bwrap sandbox, whose mount-namespace view of the repo conflicts
  with git's worktree identity check and breaks worktree creation. Off by default; only added
  if accepted during setup.
- **Skill Source Trust**: if skill sync is accepted, `~/.claude/skills` is wiped and
  repopulated from the source(s) you name, on every container start. The wipe is scoped to
  that one fixed directory inside this container — it can't reach anything else. The real
  trust decision is the source itself: naming one means trusting its skills unattended, with
  no per-skill review step, every time the container starts. Off by default.

See [ADR-0007](../setup-devcontainer/docs/adr/0007-agent-trust-hub-audit-response.md) for the
full reasoning and the alternatives rejected.

## 1. Detect the base devcontainer

```bash
test -f .devcontainer/devcontainer.json && test -f .devcontainer/post-create.sh && echo ok
```

If either file is missing, **refuse** — tell the user to run
`setup-devcontainer` first, and stop. Do not generate or patch anything.

## 2. Ask this tool's options

- **YOLO alias**: add `claude-yolo` for fast, unattended iteration? Uses
  classifier-based auto permission mode, not `--dangerously-skip-permissions`
  (which breaks `--worktree` — see the template's own comment for why).
- **Automatic skill sync**: sync AI-agent skills into `~/.claude/skills`
  automatically on every container start? Two ready-made suites are
  available — `mattpocock/skills` and `ken-guru/skills` — plus any
  individually named skill in `owner/repo/skill-name` form. Validate each
  individually-named pick live before rendering anything:
  `npx -y skills add <source> --list`; re-ask on a typo or unresolved source
  rather than rendering a broken command. **Only naming a source you trust
  matters here** — this installs and re-syncs that source's skills
  unattended, with no per-skill review step, and an installed skill's
  instructions can influence what the agent does inside the container.

## 3. Patch `post-create.sh`

Append [templates/install-block.sh](templates/install-block.sh) verbatim,
keyed on its own first line as the idempotency marker:

```bash
scripts/patch-if-absent.sh append .devcontainer/post-create.sh "# --- Claude Code ---" templates/install-block.sh
```

(`scripts/patch-if-absent.sh` is the base skill's shared, base-owned
primitive — reference it from `setup-devcontainer/scripts/`, don't copy it.)

If YOLO alias was accepted, also append
[templates/yolo-alias-block.sh](templates/yolo-alias-block.sh), under its own
independent marker so declining now and adding it later never re-triggers
the install:

```bash
scripts/patch-if-absent.sh append .devcontainer/post-create.sh "# --- Claude Code yolo-alias ---" templates/yolo-alias-block.sh
```

## 4. Patch the Network Manifest — unconditional, every run

Ensure Claude Code's own `.claude.networkAllowlist` entry exists in
`.devcontainer/network-manifest.json`, regardless of any answer above — this
is what lets an opted-in network firewall (a separate, base-owned layer)
allow the hosts Claude Code itself needs:

```bash
scripts/patch-json-array-if-absent.sh .devcontainer/network-manifest.json .claude.networkAllowlist templates/network-manifest-entries.json
```

## 5. Patch `post-start.sh` (only if skill sync was accepted)

Render [templates/post-start-block.sh](templates/post-start-block.sh) with
`{{SKILLS_SOURCES_COMMANDS}}` substituted to one `npx -y skills add <source>
--skill '*' -a claude-code -y --copy -g` line per whole-suite source (or
`--skill '<name>'` per individual pick), then append it:

```bash
scripts/patch-if-absent.sh append .devcontainer/post-start.sh "# --- Claude Code skill-sync ---" <rendered-block>
```

## 6. Patch `README.md`

Append [templates/readme-bullet.md](templates/readme-bullet.md) to the end of
the file — the base skill positions `## Installed CLI Tools` as `README.md`'s
last section, so a plain append always lands in the right place:

```bash
scripts/patch-if-absent.sh append .devcontainer/README.md "- Claude Code" templates/readme-bullet.md
```

## 7. Verify

Re-run every `patch-if-absent.sh`/`patch-json-array-if-absent.sh` command
from steps 3–6 a second time — each should produce zero diff. No dedicated
verify script exists for this skill; this re-run *is* the verification.

## 8. Report next steps

From ken-guru/skills, MIT.

More from ken-guru/skills