v2.0.0

Changelog

• 39346ab: Merge pull request #10 from cirolini/eval/real-models (@cirolini)
• 808f021: Merge pull request #17 from cirolini/main (@crnvl96)
• 71f3d5a: Merge pull request #18 from crnvl96/docs/doc-comments (@crnvl96)
• 69d4be0: Merge pull request #3 from crnvl96/main (@crnvl96)
• 753e08b: Merge pull request #4 from cirolini/docs/doc-comments-adjusts (@crnvl96)
• ecaa70f: Merge pull request #5 from cirolini/phase-0/modernize (@cirolini)
• 7980ed8: Merge pull request #6 from cirolini/phase-1/providers (@cirolini)
• 1c41270: Merge pull request #7 from cirolini/phase-2/risk (@cirolini)
• 5712240: Merge pull request #8 from cirolini/phase-3/guardrail (@cirolini)
• d1bc4e8: Merge pull request #9 from cirolini/phase-4/release (@cirolini)
• 1d71a79: Update README.md (@crnvl96)
• 459db9b: Update README.md (@crnvl96)
• 2bf0897: Update README.md (@crnvl96)
• 69ff8e5: build(docker): multi-stage build, drop the script.sh wrapper (Claude Opus 5 noreply@anthropic.com)
• 32d22fe: build: release with GoReleaser instead of by hand (Claude Opus 5 noreply@anthropic.com)
• b2e8420: build: ship docs/ and examples/README.md in the release archives (Claude Opus 5 noreply@anthropic.com)
• f8cc026: doc: adjusts on documentation comments (@crnvl96)
• f17d044: docs(results): first measurements against real models (Claude Opus 5 noreply@anthropic.com)
• 31d4c79: feat(cli): add --provider, --model, --base-url and --lang (Claude Opus 5 noreply@anthropic.com)
• 35cb925: feat(cli): add explain check with risk-based exit codes (Claude Opus 5 noreply@anthropic.com)
• 93c2a1f: feat(cli): lead with the verdict, add --json (Claude Opus 5 noreply@anthropic.com)
• bdd98bf: feat(cli): rebuild the command around the Provider interface (Claude Opus 5 noreply@anthropic.com)
• 35d5ff9: feat(config): per-provider model defaults and language selection (Claude Opus 5 noreply@anthropic.com)
• 1f015a4: feat(config): resolve settings from config.toml, .explainrc and env (Claude Opus 5 noreply@anthropic.com)
• 3b25584: feat(examples): a Claude Code hook and a ?? shell prefix (Claude Opus 5 noreply@anthropic.com)
• c47e8f7: feat(llm): add a Provider interface and an OpenAI adapter (Claude Opus 5 noreply@anthropic.com)
• c8b652b: feat(llm): stream output and add the Anthropic adapter (Claude Opus 5 noreply@anthropic.com)
• 9c046be: feat(prompt): ask for touches, reversibility and an alternative (Claude Opus 5 noreply@anthropic.com)
• 56dbaff: feat(prompt): fence the command as untrusted data (Claude Opus 5 noreply@anthropic.com)
• a037ad6: feat(prompt): write the explanation in pt-BR on request (Claude Opus 5 noreply@anthropic.com)
• 1480aaf: feat(report): render verdicts for a terminal and for programs (Claude Opus 5 noreply@anthropic.com)
• 230aca0: feat(risk): deterministic rules, held as data (Claude Opus 5 noreply@anthropic.com)
• fb949d6: feat(risk): parse shell lines instead of matching text (Claude Opus 5 noreply@anthropic.com)
• 997f347: feat(risk): read severities back from JSON (Claude Opus 5 noreply@anthropic.com)
• b6f6b21: feat(risk): six rules for gaps the eval surfaced (Claude Opus 5 noreply@anthropic.com)
• ebab9da: refactor: replace the v1 packages with cmd/explain (Claude Opus 5 noreply@anthropic.com)
• 6f50f7c: test(eval): correct seven labels after a blind cross-check (Claude Opus 5 noreply@anthropic.com)
• 982a64e: test(eval): measure real models, blind and with the rules' context (Claude Opus 5 noreply@anthropic.com)
• e63e962: test(eval): re-score a saved run without asking the model again (Claude Opus 5 noreply@anthropic.com)
• cb19722: test(eval): score verdicts against labels, and publish the numbers (Claude Opus 5 noreply@anthropic.com)
• 7534d5a: test(risk): an evaluation table of 60 commands (Claude Opus 5 noreply@anthropic.com)

Verify

sha256sum -c checksums.txt --ignore-missing

Every artifact carries a build provenance attestation. Verify it with:

gh attestation verify explain_2.0.0_linux_amd64.tar.gz --repo cirolini/explain