Architecture Vision (living)

This directory holds the current and future architecture vision for Spec Kitty — forward intent that may still change. It is part of the living architecture at the top of architecture/ (see ../README.md for the boundary and decay rules).

What belongs here

• The synthesized "where the architecture is going" narrative for the active era.
• Forward-looking structural intent that has not yet been ratified as an ADR.

What does NOT belong here

• Ratified decisions — those are ADRs under architecture/<version>/adr/ (immutable, era-stamped).
• Explorations / spikes — those are research notes under architecture/<version>/research/.
• Historical vision — when a vision statement is no longer current/future it is demoted into its era directory architecture/<version>/vision/ (the decay path; nothing is deleted).

Vision vs Decision vs Spike

Artifact	Meaning	Home	Mutability
Vision	Forward intent	docs/architecture/vision/ (top-level, living)	May change
Decision (ADR)	Ratified decision	architecture/<version>/adr/	Immutable, era-stamped
Spike	Exploration	architecture/<version>/research/	Versioned record

Current-era forward signal (2026-08-12)

Forward intent, corroborated by the 2026-08-12 code-health measurement (full detail: Code Quality — Working Collection). Not yet an ADR.

The 3.x north-star — a doctrine-governed, charter-activated runtime with a hardened execution model — is showing up in the measurements, not just the design:

• The stabilization cycle is converging. Coverage recovered from a mid-2026 trough of ~47% to a project-high 84%, reliability bugs cleared to 0, and duplication holds at 0.5%. The next control loop and the execution-lane / coord-primary model are built and canonical; the health signal now tracks the "hardened execution model" intent rather than lagging it.
• The remaining structural debt is scoped, not drifting. The one standing red — a SonarCloud security_rating of E from ~21 pre-existing subprocess/path findings — concentrates in the lane / coordination / sync surfaces the 3.x execution model built out, and maps almost 1:1 onto the already-planned Wave 2 / Wave 4 degod slices. The forward intent is to burn that backlog down through the degod program (with characterization tests first), not as cosmetic passes, and to keep the charter-as-sole-door and governance-honesty through-lines as the gate on new debt.
• Governance honesty holds as a design principle. "Red main is honest; CI is the release authority" (ADR 2026-07-17-1) is what lets an honest standing red coexist with a cuttable release candidate — the release posture is read from the functional gates, with the Sonar backlog tracked as known/deferred.