Changelog

The standard answers what it left open. 1.4 shipped two questions declared rather than settled, and a coherence review of the whole corpus found more of the same shape: places where two parts of the standard say different things about what is owed at a given configuration, and places where an obligation exists but the text that states it cannot be found by the reader who owes it.

Settled

• QG-3 at Foundation — the criteria are owed at every configuration, the gate stays Team+. Foundation still operates two gates
• Team staffing (11.2) — the Responsibilities row yields to 4.8, the single source for headcount

Added

• Section 9.4 — Reporting a figure — the method is disclosed with the figure; a figure computed from self-made records measures what was recorded; a self-set target carries its basis; a crossed target is escalated, not quietly moved
• 8.3 — root cause as a QG-2 criterion. The words appeared nowhere in the Standard while SENAR Core required the analysis twice
• 10.15(f) — the severity scale is documented, and states which of three scales it defines
• 13.1(e) — non-conformances are disclosed with the claim, as unimplemented SHOULD requirements already were
• 2.1 — SENAR Core is a normative reference; the Foundation rule set cannot be reconstructed without it
• 3.45 — Handoff, which carried SHALL in 6.5(c) and 7.8 and was defined nowhere

Changed

• 4.1 recognizes one case of direct modification, not two — an environment where no agent can be run. The list is dated in the text and tied to the re-evaluation trigger of 10.13
• Section 12 is informative — it states no requirement and cannot until three things are settled
• Section 10 yields to Section 11 for scope; 8.2 no longer states a level for configurations without QG-1; 5.6 scales with the Tooling row of 11.4
• Both PDF builds carry the whole corpus — 34 documents each

The standard norms its own gate. Until 1.4 the quality gates listed criteria and none of them named the effect the gate prevents, so a gate could be placed anywhere and conform everywhere — and the prohibition on bypass had no subject, because a bypass is undefinable until placement is defined.

Added

• Section 8.6 — Gate Properties — five properties every gate SHALL have: declared prevented effect, preventive placement, correspondence of the verdict to state, binding to a measured digest, fail-closed
• A declared prevented effect for every gate — QG-0 through QG-4 each name the change that does not occur while the verdict is negative
• 8.6(j) — direct modification of an artifact is a Gate Bypass and is recorded as one; a regulated exception with a mandatory record, not a prohibition
• Section 13.7 — Mechanically Checkable Clauses — what a claimant presents per property and how it is checked
• Coverage Limitations — a bound on the survey, next to the existing bound on the data: what was read, what was seen only in structure, and what was not examined at all

Changed

• Breaking for existing conformance claims — gate properties scale by configuration, and 13.6 now requires reassessment rather than recommending it. A gate placed after the effect it declares is a detective control and does not satisfy 8.6(b)
• Metric 8 derived from records — Manual Intervention Rate is computed from recorded Gate Bypasses instead of self-report, and says plainly that it measures what was recorded, not what occurred
• The producer qualifier is gone — rules no longer say "AI-generated code", because the same file changed the same way was outside the requirement whenever the change arrived by another route
• Positioning restated through properties — "first methodology" and "the missing layer" removed from every address; the claim is now checkable and says how to refute it

Open

• QG-3 at Foundation — 8.7 and Sections 3, 8.4, 8.9, 8.10 and 11.1 disagree on whether QG-3 is owed at Foundation for security-sensitive systems. This revision states the conflict rather than settling it, and carries it to the next

Added

• Rule 15: AI Output Quality Verification — three verification levels (L1 Automated, L2 Self-Review, L3 Adversarial Review) with risk-based scoping
• Metric 10: Adversarial Detection Rate (ADR) — measures CRITICAL findings per L3-reviewed task (renamed from Vibe Code Rate)
• S5.5 Structured Tool Protocol — protocol-agnostic requirement for agent-platform interaction
• S5.6 Agent Dispatch and Execution Isolation — risks, mandatory practices, canonical pattern
• S5.7 Federation — scaling SENAR across projects: independence, coordination, knowledge routing, federated metrics
• 23-item AI Review Checklist — expanded from 13 items with 3 priority tiers; latent defect patterns from adversarial audits
• Code Standards Template — evidence-based template for organizational code standards
• 8 new term definitions: Adversarial Review, Agent Dispatch, Latent Defect, ADR, and more
• KCR calibration guidance, no-epoch filtering policy, multi-session metrics disclaimer

Changed

• Rule 12 Context Hygiene — upgraded to SHALL for credentials/PII at ALL configurations
• Stack neutrality — removed all vendor/product names from normative text
• QG-2: “types clean” → “static analysis passes”
• Configuration counts: 15 rules (was 14), 10 metrics (was 9)

Reviewed

• 3 rounds of adversarial review by 9 independent reviewers (methodology critic, security expert, practitioner)
• All CRITICAL findings resolved including stack neutrality, human-in-the-loop, federation security

Added

• S5 Agent Instrumentation — new section: control levels, agent profiles, script governance, programmatic interface, portability
• Rule 14: Script Change Management — version-controlled, reviewed, rollbackable
• Guide: Agent Configuration in Practice — new chapter with examples for Claude Code, Codex, Cursor
• Context Architect role expanded: agent profile design, script management
• Tooling Requirements: agent configuration management section
• Changelog (this document)
• Site: mobile navigation, changelog page

Changed

• Sections renumbered: S5→S6, S6→S7, ... S12→S13
• All cross-references updated

Fixed

• Navigation 404 on Standard link
• Literary Russian improvements

Added

• Requirements hierarchy: BR → SR → TR
• QG-0 strengthened: requirement traceability
• QG-1: Requirements Gate (Team+ only)
• Rule 10: Requirement Traceability
• Rule 11: Code Documentation as Context
• Rule 12: Context Hygiene
• Rule 13: AI Model Governance
• Guide chapters: Quick Start, Requirements Engineering, Legacy Adoption, Worked Example, Tool Integration
• Conformance clause (self-declared, peer, audited)
• Enterprise roles (Portfolio Manager, Chief Supervisor)

Initial Release

• 12 sections, 5 roles, 5 units of work
• 7 procedures, 5 quality gates, 9 metrics, 11 rules
• 3 configurations: Essential, Team, Enterprise
• Guide: 6 chapters
• Reference: glossary, scaling, efficiency, governance, tooling