kube-coder: Skill for Claude Code

kube-coder preflight (run CI locally)

Mirrors .github/workflows/ci.yml so you catch failures before pushing instead of waiting on a CI round-trip.

Run kc-scope-pr first. This skill runs whole suites; that one tells you which suites your change actually reaches, and what it left untested. A green preflight over a suite that never touches your diff is not evidence. Run the subset named by $ARGUMENTS (default all). Report a concise PASS/FAIL summary per section at the end.

Run from the repo root (/home/dev/kube-coder).

Prereqs (install once, idempotent)

export PATH="$HOME/.local/bin:$PATH"
# helm (if missing)
command -v helm >/dev/null || {
  cd /tmp && curl -fsSL https://get.helm.sh/helm-v3.14.4-linux-amd64.tar.gz | tar xz \
    && install -m755 linux-amd64/helm ~/.local/bin/helm; cd -; }
# helm-unittest plugin (if missing)
helm plugin list 2>/dev/null | grep -q unittest \
  || helm plugin install https://github.com/helm-unittest/helm-unittest.git --version v0.5.2

helm — lint + template + unit tests

export PATH="$HOME/.local/bin:$PATH"
helm lint charts/workspace/ -f charts/workspace/tests/test-values.yaml
helm template test-ws charts/workspace/ -f charts/workspace/tests/test-values.yaml > /dev/null
helm unittest charts/workspace/
# controller chart too, if you touched it:
# helm lint charts/workspace-controller/ -f charts/workspace-controller/tests/test-values.yaml
# helm unittest charts/workspace-controller/

python — server.py unit + integration tests

make python-tests
# Prefer the make target: it points KC_FEED_DIR / KC_PUSH_DIR / KC_PROVIDER_KEYS_FILE
# at a temp dir, so a test can't write this workspace's live Feed, page its phone
# (#685), or read/overwrite the API keys you set in Settings. By hand:
# t=$(mktemp -d) && cd charts/workspace && KC_FEED_DIR=$t/feed KC_PUSH_DIR=$t/push \
#   KC_PROVIDER_KEYS_FILE=$t/provider-keys.json \
#   python3 -m unittest discover -s tests -p '*_test.py'