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.