Skip Build and Tests for website-only pushes and run Benchmarks only on product source

Build and Tests ignores pushes and pull requests that change only website inputs.
Benchmarks starts automatically only when src/** or the root build/package inputs
change; test, harness, documentation and website changes use manual dispatch.
Verified: workflow YAML parses; RepositoryGovernance verify passes.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Author: KSemenenko

.github/workflows/benchmarks.yml

@@ -2,6 +2,13 @@ name: Benchmarks
 on:
   push:
     branches: [main]
+    # Automatic runs follow KeyLoad product source only; tests, harness, docs and site use manual dispatch.
+    paths:
+      - 'src/**'
+      - 'Directory.Build.props'
+      - 'Directory.Build.targets'
+      - 'Directory.Packages.props'
+      - 'global.json'
   workflow_dispatch:
 permissions: {contents: read, actions: read}
 concurrency:

.github/workflows/build-and-tests.yml

@@ -2,7 +2,23 @@ name: Build and Tests
 on:
   push:
     branches: [main]
+    paths-ignore:
+      - 'site/**'
+      - 'scripts/Features/BenchmarkComparisons/site-*'
+      - 'scripts/Features/BenchmarkComparisons/composite-site-*'
+      - '.github/workflows/website.yml'
+      - '.github/workflows/Features/BenchmarkComparisons/QualifySite/**'
+      - '.github/workflows/Features/BenchmarkComparisons/BuildIsolatedSite/**'
+      - '.github/workflows/Features/BenchmarkComparisons/DeploySite/**'
   pull_request:
+    paths-ignore:
+      - 'site/**'
+      - 'scripts/Features/BenchmarkComparisons/site-*'
+      - 'scripts/Features/BenchmarkComparisons/composite-site-*'
+      - '.github/workflows/website.yml'
+      - '.github/workflows/Features/BenchmarkComparisons/QualifySite/**'
+      - '.github/workflows/Features/BenchmarkComparisons/BuildIsolatedSite/**'
+      - '.github/workflows/Features/BenchmarkComparisons/DeploySite/**'
   workflow_dispatch:
   workflow_call:
 env:

docs/Features/BenchmarkComparisons/SessionExecution.md

@@ -331,7 +331,10 @@ measured workload reports do not close that independent gate.
 
 ## Runs and results
 
-A push to main automatically runs the complete set of 32 measurement jobs. A manual run
+A push to main that changes KeyLoad product source (`src/**` or the root
+`Directory.Build.*`, `Directory.Packages.props` and `global.json` build inputs)
+automatically runs the complete set of 32 measurement jobs; pushes that change only
+tests, the benchmark harness, documentation or the website do not start it. A manual run
 chooses the complete set or only the build/harness contracts. Preflight, cancellation,
 malformed-input, permission and provenance checks are TUnit contracts of Benchmarks
 and do not multiply the measurement matrix. The selected load profiles use the same

docs/implementation/website.md

@@ -12,7 +12,8 @@ that change the site or its build, qualification and publication inputs (an expl
 path filter, so database-only pushes do not start it) and on manual dispatch. The
 final job of Benchmarks dispatches it after a successful producer run. Build and
 Tests (`build-and-tests.yml`) builds the complete solution and runs ordinary project
-tests; its result is not a dependency of Website. Benchmarks produces authenticated
+tests; it ignores pushes and pull requests that change only these website inputs, and
+its result is not a dependency of Website. Benchmarks produces authenticated
 JSON. Release remains the prepared manual product-release workflow.
 
 Website selects the latest ready authenticated Benchmarks aggregate. With data,