v0.2.0

What's Changed

• Add reproducibility evidence for advertised profiles by @DavidBakerEffendi in #86
• Add location-level precision, recall, and range-quality metrics by @DavidBakerEffendi in #87
• Add frozen real-project evaluation protocol by @DavidBakerEffendi in #88
• Capture real-project evaluation population by @DavidBakerEffendi in #92
• Add reviewed real-project-v1 evaluation corpus by @DavidBakerEffendi in #94
• Freeze initial real-project-v1 evaluation slice by @DavidBakerEffendi in #95
• Preregister real-project-v2 evaluation slice by @DavidBakerEffendi in #96
• Add published JVM semantic-pack benchmarks by @DavidBakerEffendi in #97
• Fix Bifrost batch scan budgeting by @DavidBakerEffendi in #100
• Resolve Bifrost type-alias search hits by @DavidBakerEffendi in #102
• Benchmark workspace generated behavior models by @DavidBakerEffendi in #101
• Fix resilient Bifrost benchmark execution by @DavidBakerEffendi in #104
• Add blinded agent review protocol by @DavidBakerEffendi in #105
• Enhancement: record blinded agent-panel pilot by @DavidBakerEffendi in #106
• Enhancement: support per-case agent review sessions by @DavidBakerEffendi in #108
• Complete real-project-v1 agent panel review milestones by @DavidBakerEffendi in #109
• Promote real-project-v1 evaluation publication by @DavidBakerEffendi in #110
• Prepare UsageBench v0.2.0 release by @DavidBakerEffendi in #111
• Fix Git support in reference runtimes by @DavidBakerEffendi in #112
• Fix directory symlinks in LSP source copies by @DavidBakerEffendi in #113
• Fix evaluation report publication scope by @DavidBakerEffendi in #114

Full Changelog: v0.1.0...v0.2.0