Candidate: Signalproof decomposition for new Workbench Core

[docreo]

Purpose

Introduce signalproof-decomposition as a governed Candidate Skill derived from the OpenAI Codex architecture teardown and the owner's decision to use decomposition as the new Signalproof Workbench Core starting pattern.

Direction

• preserve proven legacy Workbench behavior and integrate through compatibility boundaries;
• target Hermes-style operational flexibility without adopting Hermes branding;
• keep agent identity separate from model/provider identity;
• preserve Signalproof Governor authority, sessions/bots distinctions, evidence surfaces, local-first objectives, and established Workbench UX;
• preserve the Signalproof black/charcoal/red product language;
• generalize decomposition as a decision SOP for heavy products including Media Studio, Music Studio, Video Studio, and Greenlight;
• avoid automatic microservice fragmentation for small cohesive tools.

Governance

This PR is Candidate / Not Active. It does not activate the skill, deploy code, or authorize canonical Build Ledger chronology. A corresponding private Build Ledger record is staged as noncanonical pending ledger-head verification.

Source reference

Architecture study used the official openai/codex source at fixed commit 28327355b861ab6cc76b01c7248663eb1be440cf as evidence/reference only. Signalproof owns the resulting architecture and implementation.