diff --git a/.github/copilot-instructions.md b/.github/copilot-instructions.md
new file mode 100644
index 0000000000..de4b9077c0
--- /dev/null
+++ b/.github/copilot-instructions.md
@@ -0,0 +1,31 @@
+# Copilot Code Review ΓÇö Global Rules
+
+## Review Style
+- Be concise and factual.
+- Use short bullet points.
+- No praise, no summaries, no speculation.
+- Do not explain obvious language syntax.
+
+## Review Scope Guardrails
+- Comment only on changed lines or directly affected code.
+- If context is insufficient, say: "Insufficient context to assess."
+- Do NOT infer intent or future design.
+
+## Comment Format
+Use one of the following prefixes only:
+- **[BLOCKER]** ΓÇö Correctness, safety, resource leak, or API stability issue.
+- **[ISSUE]** ΓÇö Likely bug or deviation from established patterns.
+- **[SUGGESTION]** ΓÇö Improvement that is not blocking.
+- **[QUESTION]** ΓÇö Clarification needed from the author.
+
+## Output Constraints
+- Max 2 comments per concern.
+- Group related observations into a single comment.
+- Prefer actionable phrasing: "Missing `Close()` call on X" not "Consider adding cleanup."
+
+## Prohibited Behavior
+- Do not guess architectural intent.
+- Do not recommend refactors without a concrete defect.
+- Do not suggest changes unrelated to the diff.
+- Do not attempt to re-derive CI results.
+- If CI failures are visible in the PR, explain the failure concisely.
diff --git a/.github/instructions/v2-shim.instructions.md b/.github/instructions/v2-shim.instructions.md
new file mode 100644
index 0000000000..b65dfa5c70
--- /dev/null
+++ b/.github/instructions/v2-shim.instructions.md
@@ -0,0 +1,146 @@
+---
+applyTo:
+  - "cmd/containerd-shim-runhcs-v1/**/*.go"
+  - "cmd/containerd-shim-lcow-v2/**/*.go"
+  - "internal/hcsoci/**/*.go"
+  - "internal/uvm/**/*.go"
+  - "internal/layers/**/*.go"
+  - "internal/hcs/**/*.go"
+  - "internal/gcs/**/*.go"
+  - "internal/cow/**/*.go"
+  - "internal/resources/**/*.go"
+  - "internal/jobcontainers/**/*.go"
+  - "internal/guest/**/*.go"
+  - "internal/shim/**/*.go"
+  - "internal/controller/**/*.go"
+  - "internal/vm/**/*.go"
+---
+
+# V2 Shim — Code Review Rules
+
+Review rules for the containerd shim v2 path: shim binaries, HCS/OCI bridge,
+UVM lifecycle, resource management, guest compute service, VM controller,
+and container/process abstractions.
+
+The **primary lens** is Go conventions and best practices. The hcsshim-specific
+rules below extend — never override — standard Go guidelines.
+
+---
+
+## Go Conventions & Best Practices
+
+### Naming
+- Follow [Effective Go](https://go.dev/doc/effective_go) naming: MixedCaps, no underscores in Go names.
+- Package names are lowercase, single-word, no plurals. Avoid stutter (`hcs.HCSSystem` -> `hcs.System`).
+- Interfaces named after the method when single-method (`io.Reader`, not `io.IReader`).
+- Acronyms are all-caps (`ID`, `HTTP`, `UVM`), not `Id`, `Http`.
+
+### Exported vs Unexported
+- **If an exported symbol has no callers outside its package, unexport it.**
+- Flag new exports that are only used internally — keep the API surface minimal.
+- Exported types, functions, and methods MUST have doc comments (`// TypeName ...`).
+- Doc comments start with the symbol name and describe *what*, not *how*.
+
+### Error Handling
+- Use `%w` for error wrapping with `fmt.Errorf`; flag bare `%v` on error values.
+- Return errors, don't panic. Panics are only for truly unrecoverable programmer bugs.
+- Check every returned error. Flag `_ = fn()` where `fn` returns an error (unless in cleanup with a log).
+- Use `errors.Is` / `errors.As` for sentinel and type checks, never `==`.
+- Prefer typed errors or sentinel errors over string matching.
+
+### Resource Management
+- Every `io.Closer` must be `Close()`d — prefer `defer obj.Close()` immediately after creation.
+- If `Close()` can fail and it matters, capture the error: `defer func() { retErr = obj.Close() }()`.
+- Every handle (`os.File`, `syscall.Handle`, `windows.Handle`) must be closed in error paths.
+- Flag resources created in a loop without per-iteration cleanup.
+
+### Concurrency
+- Always pass `context.Context` as the first parameter; never store it in a struct.
+- Goroutines must respect `ctx.Done()`. Flag goroutines without cancellation.
+- Protect shared mutable state with `sync.Mutex` or channels. Flag unprotected access.
+- Never capture loop variables in goroutine closures without rebinding (pre-Go 1.22).
+- Prefer `sync.Once` for lazy initialization over manual bool + mutex patterns.
+
+### Interfaces
+- Keep interfaces small (1-3 methods). Accept interfaces, return concrete types.
+- Define interfaces at the consumer, not the implementer.
+- Flag empty interfaces (`interface{}`/`any`) when a specific type would work.
+
+### Tests
+- Test helpers must call `t.Helper()`.
+- Use `t.Cleanup()` for teardown instead of manual defers when appropriate.
+- Use table-driven tests for repetitive cases.
+- Test names: `TestFunctionName_Condition_ExpectedResult`.
+- Use `cmp.Diff` for struct comparison; `maps.Clone()` for copying maps.
+- Use `functional` build tag for tests needing a live VM or container.
+
+### Packages & Imports
+- Group imports: stdlib, external, internal. Use `goimports` ordering.
+- No circular dependencies. Flag new cross-package imports that break layering.
+- Internal packages should not leak implementation details through exported types.
+
+---
+
+## hcsshim-Specific Rules
+
+### Resource Lifecycle (CRITICAL)
+
+- Every allocated resource MUST be registered with `r.Add(...)` or `r.SetLayers(...)`.
+- Flag any `ResourceCloser` returned from a helper that is NOT added to the resource tracker.
+- Flag resources allocated inside a loop where early `return` could skip `r.Add(...)`.
+- `CreateContainer` and similar orchestrators MUST `defer` a call to
+  `resources.ReleaseResources(ctx, r, vm, true)` on error.
+- `ResourceCloserList.Release()` releases in REVERSE order. Do not manually reorder.
+
+### HCS / UVM Object Lifecycle
+
+- Every `hcs.System` or `cow.Container` MUST be `Close()`d.
+- Every `cow.Process` MUST be `Close()`d after `Wait()` completes.
+- `uvm.CreateLCOW` / `uvm.CreateWCOW` returns a UVM that MUST be `Close()`d on error.
+- `uvm.Start()` must be called after creation; flag orphaned UVMs.
+- Flag SCSI, vSMB, vPMEM, or Plan9 mounts added to a UVM but not tracked for cleanup.
+
+### Memory & Handle Leaks
+
+- Flag `syscall.Handle` or `windows.Handle` values not closed in error paths.
+- Flag goroutines that capture a `*cow.Process` or `*hcs.System` without ensuring
+  the object outlives the goroutine.
+
+### Concurrency (hcsshim-specific)
+
+- `hcs.System` operations are NOT goroutine-safe. Flag concurrent access without sync.
+- UVM device maps (`scsiLocations`, `vsmb`, `plan9`, `vpmem`) are mutex-protected;
+  flag direct map access without holding the lock.
+
+### Package Layering
+
+Flag these violations:
+- `internal/cow` importing anything above it (must be pure abstraction).
+- `internal/hcs` importing from `internal/hcsoci` or `internal/uvm`.
+- `internal/resources` importing from shim-level packages.
+- `cmd/` directly importing `internal/hcs` instead of going through `internal/hcsoci`.
+- `internal/controller/vm` importing from `internal/hcsoci`.
+
+### VM Controller & State Machine
+
+- Every state transition MUST be validated against allowed transitions.
+- Flag direct state assignment that bypasses the transition validation function.
+- Terminal states (`stopped`, `failed`) must not allow further transitions.
+- `vmmanager.LifetimeManager` and `guestmanager.Manager` MUST handle idempotent `Stop()`/`Close()`.
+- Flag implementations that panic or return untyped errors on double-close.
+
+### LCOW Shim V2 Service
+
+- Every containerd RPC handler MUST have proper context propagation and cancellation.
+- Flag handlers that block indefinitely without respecting `ctx.Done()`.
+- Flag missing cleanup in `Delete` — all resources from `Create` must be released.
+- The shim MUST NOT leak UVM references across task boundaries.
+
+---
+
+## Review Output
+
+- Max 2 comments per concern; group related items.
+- Use **[BLOCKER]** only for resource leaks, correctness, safety, or API-breaking issues.
+- Use **[ISSUE]** for likely bugs or pattern deviations.
+- Use **[SUGGESTION]** for non-blocking improvements.
diff --git a/.github/workflows/copilot-review.yml b/.github/workflows/copilot-review.yml
new file mode 100644
index 0000000000..503d4b45ae
--- /dev/null
+++ b/.github/workflows/copilot-review.yml
@@ -0,0 +1,82 @@
+# Auto-request GitHub Copilot review on PRs that touch v2 shim code.
+#
+# Copilot review is guidance-only (non-blocking). CI remains the enforcement gate.
+# The path filter ensures Copilot is NOT invoked on unrelated PRs (docs, scripts, etc.).
+
+name: Copilot Review (V2 Shim)
+
+on:
+  pull_request:
+    paths:
+      # Shim binaries
+      - "cmd/containerd-shim-runhcs-v1/**"
+      - "cmd/containerd-shim-lcow-v2/**"
+      # Core shim internals
+      - "internal/hcsoci/**"
+      - "internal/uvm/**"
+      - "internal/layers/**"
+      - "internal/hcs/**"
+      - "internal/gcs/**"
+      - "internal/cow/**"
+      - "internal/resources/**"
+      - "internal/jobcontainers/**"
+      - "internal/guest/**"
+      - "internal/shim/**"
+      # VM controller & manager interfaces
+      - "internal/controller/**"
+      - "internal/vm/**"
+
+permissions:
+  contents: read
+  pull-requests: write
+
+jobs:
+  copilot-review:
+    if: github.event.pull_request.draft == false
+    runs-on: ubuntu-latest
+    steps:
+      - name: Request Copilot Review
+        uses: actions/github-script@v7
+        with:
+          script: |
+            const pr = context.payload.pull_request;
+
+            // Check if Copilot is already a reviewer to avoid duplicate requests.
+            const reviews = await github.rest.pulls.listRequestedReviewers({
+              owner: context.repo.owner,
+              repo: context.repo.repo,
+              pull_number: pr.number,
+            });
+            const alreadyRequested = reviews.data.users.some(
+              (u) => u.login === 'copilot-pull-request-reviewer'
+            );
+            if (alreadyRequested) {
+              core.info('Copilot review already requested ΓÇö skipping.');
+              return;
+            }
+
+            // Request Copilot as a reviewer.
+            try {
+              await github.rest.pulls.requestReviewers({
+                owner: context.repo.owner,
+                repo: context.repo.repo,
+                pull_number: pr.number,
+                reviewers: ['copilot-pull-request-reviewer'],
+              });
+              core.info(`Copilot review requested on PR #${pr.number}`);
+            } catch (err) {
+              // Non-fatal: Copilot may not be enabled on the repo.
+              core.warning(`Could not request Copilot review: ${err.message}`);
+            }
+
+            // Add a label to indicate Copilot review was requested.
+            try {
+              await github.rest.issues.addLabels({
+                owner: context.repo.owner,
+                repo: context.repo.repo,
+                issue_number: pr.number,
+                labels: ['copilot-review'],
+              });
+            } catch (err) {
+              core.info(`Could not add label: ${err.message}`);
+            }