Refactor cleanup branch integration

.env

@@ -0,0 +1 @@
+export PATH="$PWD/bin:$PATH"

.github/workflows/ci-pr.yml

@@ -0,0 +1,33 @@
+name: PR CI
+
+on:
+  pull_request:
+    branches:
+      - main
+    types:
+      - opened
+      - reopened
+      - synchronize
+
+jobs:
+  checks:
+    name: Test and Build
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v4
+
+      - name: Set up Go
+        uses: actions/setup-go@v5
+        with:
+          go-version: '1.24.x'
+          cache: true
+
+      - name: Run tests
+        run: make test
+
+      - name: Build CLI
+        run: make hind-cli
+
+      - name: Build Images
+        run: ./bin/hind build all

.gitignore

@@ -1,3 +1,8 @@
 .vscode
 /bin/
 TODO
+
+skills-lock.json
+.claude/skills/golang-pro/
+.claude/worktrees/
+.worktrees/

.team/backlog.md

@@ -0,0 +1,23 @@
+# Team Backlog
+
+Closed items: `.team/done/`
+
+|  ID   | Title | Type | Priority | Size | Source | Spec |
+|-------|-------|------|----------|------|--------|------|
+| W-026 | Add GitHub Actions CI for pre-merge checks | maintenance | P1 | S | User | `W-026.md` |
+| W-033 | Fix top-priority CLI correctness bugs (`get`/`rm` missing cluster, `stop` partial failure exits) | maintenance | P1 | M | QA audit | `W-033.md` |
+| W-017 | Close `hind-stop.feature` behavior gaps | feature | P2 | L | B-015 audit | `W-015.md` |
+| W-019 | Enforce `default-cluster.feature` profile-selection contracts | feature | P2 | M | B-015 audit | `W-015.md` |
+| W-022 | Refactor `pkg/build/release` for a more ergonomic interface | maintenance | P2 | M | User | `W-022.md` |
+| W-034 | Fix low-priority CLI behavior/UX bugs from QA (`list`/`start`/`build`/`set`) | maintenance | P3 | M | QA audit | `W-034.md` |
+| W-020 | Fix all `go` test files that use duplicate mocks | maintenance | P3 | S | User | `W-020.md` |
+| W-021 | Address B-013 staff review additional observations | maintenance | P3 | S | B-013 review | `W-021.md` |
+| W-023 | Allow users to pass arbitrary package versions to `hind build` | feature | P3 | M | User | `W-023.md` |
+| W-024 | Allow users to pass arbitrary package versions to `hind start` | feature | P3 | M | User | `W-024.md` |
+| W-025 | Publish container images to an OCI registry on version update | feature | P4 | XL | User | `W-025.md` |
+| W-027 | Add ID value to cluster nodes `pkg/cluster/types.go` | maintenance | P3 | S | User | `W-027.md` |
+| W-028 | Migrate `*_test.go` test cases to `stretchr/testify` | maintenance | P4 | L | User | - |
+| W-029 | Add ingress controller for routing traffic to the internal network | feature | P4 | XL | User | - |
+| W-030 | Build and publish releases to brew for macos install | feature | P3 | L | User | - |
+| W-031 | Add open subcommand to open the web ui of a component | feature | P3 | M | User | - |
+| W-032 | Add login subcommand to exec into an interactive shell in a node | feature | P3 | M | User | - |

.team/bugs.md

@@ -0,0 +1,162 @@
+# Bugs
+
+Closed items: `.team/done/bugs/`
+
+---
+
+<!-- Template:
+## W-xxx — Title
+
+**Severity:** P1 / P2 / P3
+**Status:** open / in-progress / resolved
+**Source:** where it came from
+
+**Repro steps:**
+1. ...
+
+**Expected:** ...
+**Actual:** ...
+
+**Spec:** `W-xxx.md` (if one exists)
+-->
+
+---
+
+## BUG-001 — `hind get` succeeds silently when cluster has no config file on disk
+
+**Severity:** P2
+**Status:** open
+**Source:** QA audit of assigned commands (get, list, stop, set profile)
+
+**Root cause file:** `pkg/cluster/manager.go:325-331` (`LoadPersistedConfig`)
+
+**Repro steps:**
+1. Ensure no cluster named "ghost" has ever been created (no config file at `~/.config/hind/cluster/ghost/cluster.json`).
+2. Run `hind get ghost`.
+3. Observe exit code and output.
+
+**Expected:** Command exits non-zero with a "cluster not found" error message.
+
+**Actual:** `LoadPersistedConfig` (called by `Manager.Get`) falls through the `!m.ConfigFileExists()` branch without returning an error because `m.config.Name` is non-empty (it was set by `cluster.New` from the supplied arg). The manager then calls `provider.InspectNetwork` and `provider.InspectContainer` using the freshly-synthesised default config, both return `nil` (no such resources), and `Get` returns an empty `ClusterInfo` with no error. `get.runE` renders a table showing `Status: N/A`, `Network: ` (empty), and exits 0. The user receives no indication that the cluster does not exist.
+
+**Contrast with `hind stop`:** `stop.runE` explicitly calls `clusterMgr.ConfigFileExists()` before proceeding and returns an error if false. `get.runE` has no equivalent guard.
+
+---
+
+## BUG-002 — `hind list` swallows `tabwriter.Flush` error
+
+**Severity:** P3
+**Status:** open
+**Source:** QA audit of assigned commands (get, list, stop, set profile)
+
+**Root cause file:** `pkg/cmd/hind/list/list.go:110`
+
+**Repro steps:**
+1. Any `hind list` invocation that reaches the table-printing path (at least one cluster exists).
+2. Observe: the return value of `w.Flush()` is discarded.
+
+**Expected:** The error returned by `w.Flush()` (e.g. a broken-pipe or closed writer) is propagated and causes `runE` to return a non-zero exit code, consistent with how `get.runE` handles `w.Flush()` at line 85-87 of `get.go`.
+
+**Actual:** `w.Flush()` is called as a bare statement with no error check (`w.Flush()` on line 110, no `if err :=` wrapper). If the flush fails the error is silently dropped and the command exits 0, potentially producing truncated output.
+
+---
+
+## BUG-003 — `hind stop` exits 0 even when containers fail to stop
+
+**Severity:** P2
+**Status:** open
+**Source:** QA audit of assigned commands (get, list, stop, set profile)
+
+**Root cause file:** `pkg/cmd/hind/stop/stop.go:115-121`
+
+**Two related issues in the same function:**
+
+**Issue A — `FailedCount > 0` branch always returns nil (line 121)**
+When one or more containers fail to stop (without `--force`), `runE` prints the partial-stop warning to `ErrOut` but returns `nil` — exit 0. Scripts cannot detect the failure.
+
+**Issue B — `--force` branch (line 115) evaluated before `FailedCount` check (line 119)**
+When `force=true` and containers failed to kill, the `if force` branch fires first and returns `nil` with "force stopped", so the `FailedCount` path is never reached at all. Individual failure messages are still printed to `ErrOut` (lines 103-105) but the exit code and final status message are incorrect.
+
+**Repro steps (Issue A):**
+1. Start a cluster with at least two containers.
+2. Arrange for one container's stop to fail (e.g. Docker daemon error on that container).
+3. Run `hind stop <cluster-name>` (no --force).
+4. Observe exit code is 0 despite partial failure.
+
+**Repro steps (Issue B):**
+1. Same setup but run `hind stop --force <cluster-name>`.
+2. Observe "force stopped" message and exit 0 — `FailedCount` path is never reached.
+
+**Expected:** Both paths should return a non-nil error when `result.FailedCount > 0`.
+
+---
+
+## BUG-004 — `hind rm` succeeds silently when cluster does not exist
+
+**Severity:** P2
+**Status:** open
+**Source:** QA audit of assigned commands (rm, start, build)
+
+**Root cause file:** `pkg/cluster/manager.go:221-276` (`Delete`)
+
+**Repro steps:**
+1. Ensure no cluster named `ghost` exists.
+2. Run `hind rm ghost`.
+3. Observe output and exit code.
+
+**Expected:** Command exits non-zero with an error such as `cluster 'ghost' does not exist`.
+
+**Actual:** Command exits 0 and prints `Cluster 'ghost' deleted successfully`. `Delete()` does not fail when config/resources are absent, so non-existent deletion is treated as success.
+
+---
+
+## BUG-005 — `hind start` never returns `StartResultAlreadyRunning`
+
+**Severity:** P3
+**Status:** open
+**Source:** QA audit of assigned commands (rm, start, build)
+
+**Root cause file:** `pkg/cluster/manager.go:72-108` (`Start`)
+
+**Repro steps:**
+1. Run `hind start mycluster`.
+2. Run `hind start mycluster` again while already fully running.
+
+**Expected:** Already-running no-op path should return `StartResultAlreadyRunning` and avoid redundant connection-info output.
+
+**Actual:** Existing-cluster path always returns `StartResultResumed`, even when reconcile has no actions. `StartResultAlreadyRunning` appears to be dead/unused.
+
+---
+
+## BUG-006 — `hind build` wrong-arg error message uses slice formatting
+
+**Severity:** P3
+**Status:** open
+**Source:** QA audit of assigned commands (rm, start, build)
+
+**Root cause file:** `pkg/cmd/hind/build/build.go:39`
+
+**Repro steps:**
+1. Run `hind build` (no args), or `hind build a b`.
+
+**Expected:** Error should report argument count cleanly (e.g. `received 0` / `received 2`) or use Cobra's default exact-args error.
+
+**Actual:** Custom message uses `%s` with `[]string`, producing messages like `accepts 1 arg, received []` or `[a b]`.
+
+---
+
+## BUG-007 — `hind set profile` writes success message to stderr
+
+**Severity:** P3
+**Status:** open
+**Source:** QA audit of assigned commands (get, list, stop, set profile)
+
+**Root cause file:** `pkg/cmd/hind/set/set.go:42`
+
+**Repro steps:**
+1. Run `hind set profile <existing-cluster>`.
+2. Redirect stderr to `/dev/null`.
+
+**Expected:** Success output should be on stdout if treated as user-facing command result.
+
+**Actual:** Success message is written to `streams.ErrOut`, so it disappears when stderr is redirected.