[TristanKruse]

Summary

Fixes #65.

• Restrict the enforced SimpleCov report to this repository's lib/**/*.rb files, including tracked but unloaded runtime files. Loaded specs, benchmark code, and development scripts cannot inflate the denominator.
• Preserve the existing minimums: 98% line and 90% branch coverage. No runtime behavior or dependencies change.
• Add an isolated subprocess regression that loads the real configuration and tooling, checks the exact runtime file set and thresholds, and does not merge its partial probe execution into the full suite's coverage.
• Document scope and platform-specific symlink skips in CONTRIBUTING.

Validation

• Before fix: the regression failed because loaded benchmark code appeared in the enforced report.
• After fix: 594 examples, 0 failures (8 Windows native-symlink skips); 261 files lint-clean.
• Real library-only coverage: 98.47% line (5294/5376), 91.40% branch (1244/1361), passing unchanged thresholds.
• Independent sibling of Fail closed on malformed Ruby with structured parse diagnostics #68 and Resolve conditional constant declarations in expanded analysis #69, based on main. No merge, version bump, or publication.

Combined validation

Locally combined #68, #69, and #70 on an unpushed verification branch without conflicts. Passed: 614 examples with zero failures (8 Windows native-symlink skips); runtime-library-only coverage 98.48% line / 91.51% branch at unchanged 98% / 90% thresholds; 267 files lint-clean; 179 documentation pages validated; both public/installed-API smoke scripts; fresh scan of 1,269 advisories with zero findings and successful health detection; all six pinned Rack/Rake/RSpec canary modes with reviewed fingerprints unchanged. All GitHub checks on each PR are green. No PR was merged and no release/version bump was performed.