RGDS v2.0.1: Repository remediation

RGDS v2.0.1

This patch release includes the P0 and P1 remediation and the intervening documentation changes since the historical v.2.0.0 tag.

Changes

• Corrected outcome descriptions, example identities, field references, and the stated scope of schema and CI checks. Removed unsupported regulatory and timing claims.
• Added implementation citation metadata and preserved the previous DOI attribution in the provenance record. The Zenodo DOI remains a reference to the independent study v1.4.
• Unified semantic checks across both validator entry points and enabled date-format validation in the single-record validator. Added nine regression tests to CI.
• Revised governance, AI-assistance guidance, reviewer checklists, role mappings, and illustrative AI review artifacts.
• Revised evaluation plans, evidence rubrics, decision extracts, the scorecard template, and internal requirements traceability. These materials do not establish field effectiveness.
• Restored the complete Apache 2.0 terms from the historical release, retaining the current copyright notice. The cumulative release also includes earlier NOTICE updates.

Compatibility and evidence

The decision schemas, decision template, and six canonical JSON records are unchanged from v.2.0.0. The validator script version is 1.0.1. Closing validator gaps can reject records that previously passed, particularly invalid dates or incomplete AI disclosures. Warning severity is unchanged; strict mode continues to treat warnings as failures.

Release validation covers all four repository Make targets, strict batch validation, strict validation of each of the six canonical records, and the nine regression tests. Additional checks cover citation metadata, schema validity, relative links, and preservation of the decision contract and historical records.

RGDS remains a working reference implementation. Passing repository checks does not establish regulatory compliance, independent research validation, or Node & Norm admission.

Historical releases and tags retain their original identities and content. This release does not assign the independent-study DOI to the implementation. Codex assisted with preparation and validation under the repository owner's authorization.

v2.0.0 — Whitepaper-Aligned Decision Governance (Breaking)

Overview

RGDS v2.0.0 aligns the repository to the RGDS whitepaper’s decision-governance model.
This release strengthens RGDS as an auditable, non-agentic decision-support system by making governance requirements enforceable through schema + semantic validation.

This is a breaking release for v1.x decision logs.

---

Key Changes

Decision Log Requirements (now enforceable)

• Mandatory options analysis (≥2 options per decision)
• Explicit evidence completeness classification:
  • complete / partial / placeholder
• Residual risk captured as a first-class decision artifact
• Explicit risk posture declaration:
  • risk_minimizing / risk_neutral / risk_accepting
• Named human accountability (owner + approvals)
• Structured AI assistance disclosure when AI is used (tool identity, purpose, human review, overrides, risk assessment)

Schema, Template, and Validators

• Updated decision log schema (JSON + YAML) and template to prevent drift
• Strengthened semantic validation:
  • options minimum enforced
  • AI disclosure fields enforced when ai_assistance.used=true
• All canonical examples pass schema + semantic validation

---

Migration Notes (v1.x → v2.0.0)

Existing v1.x logs must be updated to conform to v2.0.0:

• add options_considered (≥2)
• add risk_posture enum
• add risk_assessment + residual risk statement/items
• add evidence item completeness_state
• ensure ai_assistance is present (and fully populated if used)

---

Governance Stance (unchanged)

• AI is explicitly non-agentic
• AI never decides, approves, or accepts risk
• Human accountability remains primary and explicit

RGDS v1.4.0 — Explicit governance deltas

• Adds explicit governance fields:
• evidence_completeness (complete/partial/placeholder + author-at-risk)
• propagation_required (downstream update declaration)
• risk posture benchmarking basis
• decision authority scope + escalation path
• AI assistance trust signals (confidence band + human override)
• Updates template + docs + evaluation plan + scorecard
• Updates canonical examples 0001–0005
• README canonical references updated to include examples 0003 and 0004
• Validation: python3 scripts/validate_all_examples.py ✅

v1.3.1 — Canonical IND conditional-GO decision example

Overview

v1.3.1 adds a single canonical IND decision example that demonstrates RGDS
operating under real execution constraints.

This release is designed to answer the question:
“What does this look like in practice?”

Highlights

• Canonical IND conditional-GO decision with:
  • author-at-risk drafting
  • reviewer triage
  • publishing lock points
  • dependency and readiness tracking
• Explicit, human-governed AI support artifacts (informational only)
• Documentation updates for clarity and discoverability

What this shows

• How teams can proceed responsibly with incomplete data
• How decisions are made explicit instead of reconstructed later
• How AI can support awareness without undermining accountability

Compatibility

All examples pass strict schema validation.

v1.3.0 — IND delivery alignment

Overview

v1.3.0 aligns RGDS with real-world IND execution by making delivery and
regulatory decisions explicit, auditable, and phase-appropriate.

This release is grounded in observed IND program realities:
late-arriving data, complex dependencies, reviewer bottlenecks,
and the need to accept controlled risk without losing regulatory trust.

Highlights

• IND-aware decision log schema extensions (optional, governance-controlled)
• Explicit modeling of:
  • risk posture
  • author-at-risk drafting
  • reviewer triage
  • scope changes
  • dependency and data-readiness status
  • publishing lock points
• Role → decision → artifact matrix covering PMs, writers, regulatory,
  CMC, ops, quality, and Principal AI Business Analysts
• Updated governance and documentation
• No agent autonomy; human accountability preserved

Intended audience

• Principal AI Business Analysts
• Program and delivery leaders in regulated environments
• Regulatory, quality, and governance stakeholders

Compatibility

All existing RGDS examples remain valid and pass schema validation.

RGDS v1.2.0 — Explicit Risk Posture & Defensible Conditional Decisions

RGDS v1.2.0 strengthens decision defensibility in regulated, phase-gated workflows by making previously implicit judgment calls explicit, auditable, and condition-bound.

This release is grounded in real IND execution challenges surfaced by Syner-G practitioners and focuses on surgical schema additions rather than system redesign.

What’s new

• Explicit risk_posture
  • Forces phase-appropriate risk tolerance and trade-off rationale to be stated, not implied.
• author_at_risk_items[]
  • Formalizes placeholder drafting as a governed, owned risk with verification criteria and fallback.
• review_plan
  • Captures reviewer triage decisions (required vs optional) under time pressure.
• scope_change_events[]
  • Makes scope volatility and late discoveries auditable decision inputs.
• regulatory_interaction_decision
  • Treats pre-IND and FDA interaction strategy as a first-class decision artifact.
• Required fallback_plan for conditional_go
  • Ensures contingency planning is explicit before proceeding under uncertainty.

Canonical examples

• Conditional GO with author-at-risk drafting (execution / dependency management)
• GO with pre-IND regulatory interaction strategy (risk posture / FDA alignment)

What this is (and is not)

• ✅ Human-governed decision support
• ✅ Evidence-linked and schema-validated
• ✅ Designed for auditability and regulatory credibility
• ❌ Not an autonomous or agentic system

Compatibility

• Backward-compatible with v1.1 examples (additive schema changes only).
• All examples validated against updated schema.

Why this matters

RGDS v1.2.0 closes the gap between real execution decisions and what is typically left undocumented — scope trade-offs, reviewer routing, placeholder risk, and regulatory posture — without introducing automation risk or process overhead.

This release reinforces RGDS’s core purpose:
delivery → governance → decision confidence.

RGDS v1.1 — IND-Aligned Decision Support (Non-Agentic)

IND-aligned, human-governed decision support for phase-gated regulated workflows.
Includes regulatory interaction decisions, semantic validation, change control,
and executive-ready gate extracts.

RGDS v1 — Reference Implementation

Initial reference implementation of RGDS (Regulated Gate Decision Support).

Includes:

• Schema-validated decision log model
• Canonical GO and NO-GO decision examples
• Explicit governance and auditability model
• CI enforcement for decision completeness
• Non-agentic, human-governed design

This repository is an independent case study, not a production system.