ReproMin v0.1.0.dev7

Highlights

• repomin report validate --json now emits a versioned, privacy-safe adoption summary with oracle type, size-retention ratios, holdout counts, budget state, and explicit-environment count.
• The issue chooser includes a structured workflow-feedback template for successful, inconclusive, and blocked trials.
• The GitHub Action exposes privacy-safe oracle, retention-ratio, and payload-fingerprint outputs for downstream CI gates.
• Replay mismatch evidence includes expected and actual exit codes without exposing commands, expressions, or output.

Verification

• Main CI: https://github.com/fly1d/repomin/actions/runs/33466283646 (all jobs passed)
• Local full suite: 560 passed, 2 skipped
• Offline benchmarks: 16 passed, 0 skipped, 0 failed
• twine check, isolated wheel/sdist installs, CLI smoke checks, and pip check passed

SHA-256:

• repomin-0.1.0.dev7-py3-none-any.whl: f5856dc6edee9d858cfba2954eb07724ad0b0675ac0a2c4504337f9bc979620d
• repomin-0.1.0.dev7.tar.gz: 498c562348eff3fecbcb3cd49267c1834fdcb99c945c0a1b0455df772af9294c

ReproMin remains pre-alpha. A passing oracle is evidence for one configured failure in one recorded environment, not a correctness, production-reliability, or security guarantee.