v0.6.0 — Argument quality and judgment-writing contract

Highlights

• Adds an opt-in schema_version: "0.6" contract for argument-quality and judgment-writing records without changing the canonical 24-point score or 18-point threshold.
• Adds claim→evidence→inference→action chains, counterevidence/boundaries, Path A/B, and a five-pass writing review.
• Hardens source validation against reserved/local domains, private IPs, missing hosts, legacy numeric IPv4 spellings, and fragment-based source-count bypasses.

Verified release

• Exact tag commit: 2486d47fb326e98478a61e19841b9d393b5190d6
• 35 unit tests passed.
• sdist and wheel built from the exact tag; installed in a fresh virtual environment; CLI init/validate/score/report smoke tests passed.
• SHA256SUMS.txt and PROVENANCE.json are attached.

v0.5.2 — Citation and package discovery

What changed

• Added CITATION.cff for machine-readable citation.
• Added package classifiers, keywords, repository URLs, and release badges.
• Kept the v0.5 decision-packet schema, thresholds, and CLI behavior unchanged.

Verification

python -m pytest -q
python -B -m r2d validate examples/fictional-ai-governance-research-to-decision/decision_brief.json
python -m pip wheel . --no-deps -w dist

Verified locally on 2026-07-16: 22 tests passed, CLI validation passed, and one wheel built successfully.

v0.5.1 — Professional decision-packet review context

What changed

• Expanded the README for research, policy analysis, product discovery, AI deployment, consulting, and governance review.
• Added a four-layer explanation of decision framing, evidence, decision review, and artifact/feedback.
• Preserved the v0.5 structural score, source-declaration contract, decision-review area, and CLI behavior.

Verify locally

python -m pip install -e .
python -m unittest discover -s tests -v
r2d score examples/fictional-ai-governance-research-to-decision/decision_brief.json

This patch improves documentation and package metadata. A passing score remains structural readiness for a human decision meeting, not approval or implementation authority.

v0.5.0 — Human decision-packet contract

What changed

• Reframes the 24-point result as structural readiness for a human decision meeting, not decision approval.
• Adds a six-point decision-review area: alternatives, criteria, stakeholders, reversibility, trade-off, and pre-mortem failure.
• Requires typed artifact and feedback fields plus human-declared source checks for primary and secondary sources.
• Rejects reserved/example source domains and malformed nested input.
• Adds an AI-deployment handoff guide from prototype readiness to a separate decision packet.

Upgrade note

v0.5 decision briefs require decision_review, typed artifact and feedback objects, and source-check fields for primary/secondary claims. Start from the fictional example before migrating a v0.4 brief.

Verify locally

python -m pip install -e .
python -m unittest discover -s tests -v
r2d score examples/fictional-ai-governance-research-to-decision/decision_brief.json

Boundary

The CLI validates declared structure. It does not fetch sources, authenticate checkers, verify evidence truth, or grant authority to implement.

v0.4.1 — License metadata correction

Restores the complete SPDX-standard MIT license text so GitHub and downstream tooling can identify the repository license correctly. No scoring or CLI behavior changed; all 10 tests still pass.

v0.4.0 — First hardened public release

What changed

• Corrected the scoring model to four transparent six-point areas (24 total).
• Added explicit source tiers and distinct primary-source URL checks.
• Added init, validate, score, and report CLI commands.
• Added 10 unit tests covering schema, score invariants, vetoes, reports, and CLI behavior.
• Integrated a decision-review module with pre-mortem, assumption-ledger, and red-team templates.
• Added direct source URLs and an explicit uncalibrated-method boundary.

“Ready for decision meeting” is not approval, correctness, compliance, or implementation readiness.