Add CI benchmark tracking workflow

Author: mgomes

.github/workflows/benchmarks.yml

@@ -0,0 +1,40 @@
+name: benchmarks
+
+on:
+  push:
+    branches: [master]
+  pull_request:
+  workflow_dispatch:
+
+jobs:
+  benchmark:
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v4
+
+      - name: Set up Go
+        uses: actions/setup-go@v5
+        with:
+          go-version-file: go.mod
+
+      - name: Run benchmarks
+        run: |
+          mkdir -p .bench
+          go test ./vibes -run '^$' -bench '^BenchmarkExecution' -benchmem | tee .bench/benchmark.txt
+
+      - name: Publish benchmark summary
+        run: |
+          {
+            echo "## Benchmark Results"
+            echo ""
+            echo '```text'
+            cat .bench/benchmark.txt
+            echo '```'
+          } >> "$GITHUB_STEP_SUMMARY"
+
+      - name: Upload benchmark artifact
+        uses: actions/upload-artifact@v4
+        with:
+          name: benchmark-results
+          path: .bench/benchmark.txt

README.md

@@ -174,9 +174,13 @@ Long-form guides live in `docs/`:
 This repository uses [Just](https://github.com/casey/just) for common tasks:
 
 - `just test` runs the full Go test suite (`go test ./...`).
+- `just bench` runs the core execution benchmarks (`go test ./vibes -run '^$' -bench '^BenchmarkExecution' -benchmem`).
 - `just lint` checks formatting (`gofmt`) and runs `golangci-lint` with a generous timeout.
 - Add new recipes in the `Justfile` as workflows grow.
 
+CI also publishes benchmark artifacts via `.github/workflows/benchmarks.yml` on
+pull requests and pushes to `master`.
+
 Contributions should run `just test` and `just lint` (or the equivalent `go` and `golangci-lint` commands) before submitting patches.
 
 ## Runtime Sandbox & Limits

ROADMAP.md

@@ -340,7 +340,7 @@ Goal: improve day-to-day developer productivity and interpreter robustness.
 ### v0.19.0 Definition of Done
 
 - [ ] Tooling commands are documented and stable.
-- [ ] Performance regressions are tracked with benchmarks.
+- [x] Performance regressions are tracked with benchmarks.
 - [x] CI includes example and fuzz coverage gates.
 
 ---