feat: initial repo — custom fullsend sandbox image with yarn/corepack

Extends ghcr.io/fullsend-ai/fullsend-code:latest with corepack and yarn
pre-activated for the rhdh-plugins monorepo. Includes multi-arch GHA
build workflow mirroring upstream fullsend image CI.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Author: durandom

.github/workflows/sandbox-images.yml

@@ -0,0 +1,67 @@
+name: Sandbox Images
+
+on:
+  push:
+    tags:
+      - "v[0-9]+.[0-9]+*"
+    branches:
+      - main
+    paths:
+      - "images/code/**"
+  pull_request:
+    paths:
+      - "images/code/**"
+  workflow_dispatch:
+
+# Cancel in-progress runs for the same branch/PR.
+concurrency:
+  group: ${{ github.workflow }}-${{ github.head_ref || github.ref }}
+  cancel-in-progress: true
+
+jobs:
+  build-code:
+    runs-on: ubuntu-latest
+    permissions:
+      contents: read
+      packages: write
+    env:
+      IMAGE_NAME: ghcr.io/${{ github.repository_owner }}/rhdh-fullsend-code
+    steps:
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
+
+      - name: Log in to GitHub Container Registry
+        uses: docker/login-action@4907a6ddec9925e35a0a9e82d7399ccc52663121 # v4
+        with:
+          registry: ghcr.io
+          username: ${{ github.actor }}
+          password: ${{ secrets.GITHUB_TOKEN }}
+
+      - name: Set up QEMU
+        uses: docker/setup-qemu-action@ce360397dd3f832beb865e1373c09c0e9f86d70a # v4.0.0
+
+      - name: Set up Docker Buildx
+        uses: docker/setup-buildx-action@4d04d5d9486b7bd6fa91e7baf45bbb4f8b9deedd # v4
+
+      - name: Extract metadata
+        id: meta
+        uses: docker/metadata-action@030e881283bb7a6894de51c315a6bfe6a94e05cf # v6
+        with:
+          images: ${{ env.IMAGE_NAME }}
+          tags: |
+            type=semver,pattern={{version}}
+            type=semver,pattern={{major}}.{{minor}}
+            type=sha,prefix=
+            type=raw,value=latest,enable={{is_default_branch}}
+            type=raw,value=dev,enable=${{ github.event_name != 'pull_request' }}
+
+      - name: Build and push
+        uses: docker/build-push-action@bcafcacb16a39f128d818304e6c9c0c18556b85f # v7
+        with:
+          context: images/code
+          file: images/code/Containerfile
+          platforms: linux/amd64,linux/arm64
+          push: ${{ github.event_name != 'pull_request' }}
+          tags: ${{ steps.meta.outputs.tags }}
+          labels: ${{ steps.meta.outputs.labels }}
+          cache-from: type=gha,scope=code
+          cache-to: type=gha,mode=max,scope=code

CODEOWNERS

@@ -0,0 +1 @@
+* @durandom @kadel

README.md

@@ -0,0 +1,76 @@
+# rhdh-fullsend
+
+Custom fullsend sandbox images for the RHDH team's agent infrastructure.
+
+## Why this repo exists
+
+The upstream [fullsend-code](https://github.com/fullsend-ai/fullsend) sandbox
+image ships with Go, Python, and shell tooling but no JavaScript package
+manager. The rhdh-plugins monorepo (23 workspaces, yarn) requires yarn to run
+tests, linting, and OpenSpec validation. Without it baked into the image, agents
+spend 10-15 minutes bootstrapping corepack/yarn on every run — and the
+workaround (a `host_files`-mounted shell script) is fragile.
+
+This repo builds a single image that extends `fullsend-code:latest` with
+corepack and yarn pre-activated.
+
+## Image
+
+```
+ghcr.io/fullsend-ai/fullsend-code:latest   (upstream)
+  └── ghcr.io/redhat-developer/rhdh-fullsend-code:latest   (this repo)
+```
+
+| What's added | Why |
+|-------------|-----|
+| `corepack enable` | `/usr` is read-only in the sandbox — can't enable at runtime |
+| `corepack prepare yarn@stable` | Pre-downloads yarn binary, zero cold-start |
+| `/usr/local/bin/yarn` wrapper | Git hooks (husky) run in subprocesses without the agent's PATH |
+
+## Tags
+
+| Tag | When | Use |
+|-----|------|-----|
+| `latest` | Push to `main` | Production — harness configs reference this |
+| `dev` | Any non-PR build | Testing and CI |
+| `X.Y.Z` | Tag push `v*` | Immutable release pin |
+| `X.Y` | Tag push `v*` | Floating minor for auto-patch |
+| `<sha>` | Every non-PR build | Debugging and rollback |
+
+PRs build but don't push (validation only).
+
+## Usage
+
+Reference in your fullsend harness config:
+
+```yaml
+# .fullsend/customized/harness/code.yaml
+image: ghcr.io/redhat-developer/rhdh-fullsend-code:latest
+```
+
+This replaces the `sandbox-yarn-setup.sh` + `host_files` workaround.
+
+## Local build
+
+```bash
+podman build -t rhdh-fullsend-code:local \
+  -f images/code/Containerfile images/code/
+```
+
+## Adding more tools
+
+Edit `images/code/Containerfile`. Follow the upstream pattern:
+- Pin versions via `ARG`
+- Verify checksums with `sha256sum -c` for binary downloads
+- Keep `USER sandbox` as the last line
+
+## Upstream tracking
+
+This image inherits everything from `fullsend-code:latest`. When upstream
+updates their base image (Claude Code, gitleaks, Go, etc.), our image picks
+it up automatically on the next rebuild. Pin `BASE_IMAGE` to a specific
+digest if you need reproducibility:
+
+```dockerfile
+ARG BASE_IMAGE=ghcr.io/fullsend-ai/fullsend-code@sha256:abc123...
+```

images/code/Containerfile

@@ -0,0 +1,48 @@
+# Containerfile — RHDH custom code agent sandbox image.
+#
+# Extends the upstream fullsend-code image with tools needed by the
+# rhdh-plugins monorepo:
+#   - corepack enabled with yarn pre-activated (avoids 10-15 min
+#     bootstrap inside read-only /usr sandbox)
+#   - Node.js is already in the base image (installed by Claude Code)
+#
+# This image eliminates the need for the sandbox-yarn-setup.sh
+# host_files workaround in harness configs.
+#
+# Build (native arch):
+#   podman build -t rhdh-fullsend-code:local \
+#     -f images/code/Containerfile images/code/
+#
+# Build (cross-platform):
+#   podman build --platform linux/arm64 -t rhdh-fullsend-code:arm64 \
+#     -f images/code/Containerfile images/code/
+
+ARG BASE_IMAGE=ghcr.io/fullsend-ai/fullsend-code:latest
+FROM ${BASE_IMAGE}
+
+USER root
+
+# ---------------------------------------------------------------------------
+# corepack + yarn — the sandbox filesystem policy makes /usr read-only,
+# so `corepack enable` fails at runtime. Pre-enable it here and
+# pre-download yarn so agents get yarn on PATH with zero startup cost.
+#
+# COREPACK_HOME is set to a writable location that persists across
+# the sandbox session. The shim symlinks land in /usr/local/bin
+# (writable during build, read-only at runtime — which is fine,
+# they're already there).
+ENV COREPACK_HOME=/usr/local/share/corepack
+RUN mkdir -p "$COREPACK_HOME" \
+    && corepack enable \
+    && corepack prepare yarn@stable --activate \
+    && yarn --version
+
+# ---------------------------------------------------------------------------
+# Wrapper for git hooks (husky) — hooks run in subprocesses without
+# the agent's PATH modifications. This wrapper in /usr/local/bin
+# (on default PATH) ensures `yarn lint-staged` etc. work.
+RUN printf '#!/bin/bash\nexport COREPACK_HOME=/usr/local/share/corepack\nexec /usr/bin/corepack yarn "$@"\n' \
+      > /usr/local/bin/yarn \
+    && chmod +x /usr/local/bin/yarn
+
+USER sandbox