v1.2.0

1.2.0 - 2026-May-16

This release compiles iterative improvements from practical AI-RPI Protocol usage: stronger orchestration, better product-shaping workflows, cleaner persistent context, and a more deliberate release process.

Added

• Mission Control orchestration model: main agent now acts as Mission Control — coordinating specialist subagents for research, implementation, verification, and review instead of doing everything in one pass.
• PRD-driven execution: non-trivial work is decomposed into Ralph-inspired PRDs with built-in verification, each implemented and verified by separate subagents.
• New skills: explore-options, check-project-foundations, write-implementation-plan, revise-implementation-plan, track-delivery, write-product-definition-brief.
• Product collaboration stances: commercial-mentor, product-designer, product-engineer personas for product-facing work.
• Product Definition Brief mode: specialized product flow inside Discuss mode with commercial-mentor persona, honest idea validation, and routing to foundations check or implementation planning.
• Project Foundations Check: lightweight PRD-readiness pass for idea-born public-facing products before implementation planning.
• Release playbook: core/system/release-playbook.md — end-to-end release runbook for preparing, validating, and publishing releases.
• Autonomy posture section in README — explicit exoskeleton-by-default, authorized-autonomy-only stance.
• Memory templates: decisions, lessons, metrics-log, research-cache, and session-state templates moved to templates/project-info-generation/.
• New templates: product-definition-brief-template.md, project-foundations-check-template.md.

Changed

• Skills index overhauled: Mission Control auto-trigger routing table, explicit skill-to-phase mapping, and expanded core skill set.
• Planning phase expanded: Ralph-inspired PRD decomposition guidance, tech-spec integration, and implementation plan gates strengthened.
• Implementation skill hardened: subagent delegation, PRD-scoped execution, and verification handoff formalized.
• Verification skill expanded: validation, acceptance-criteria mapping, and confidence calibration tightened.
• Packaging skill expanded: reviewer-ready output with delivery summary, PR body generation, and code review offer.
• PRD templates enhanced: all three sizes (micro, condensed, full) gained verification criteria sections and clearer acceptance mapping.
• Tech-spec templates enhanced: both sizes gained structured sections for constraints, API contracts, and migration notes.
• Core rules updated: phases, memory-handling, first-interaction, context-attention, topic-drift, formatting, input, output, self-check, self-improvement, think-first, and token-discipline rules refined for Mission Control and PRD-driven execution.
• Persistent project context restructured: three explicit surfaces (protocol, project-info, user-preferences) with .gitignore guidance for user-preferences.
• Setup lifecycle updated: bootstrap, upgrade, and health-check guidance refined.
• Project-info generation templates updated: file structure, loading guide, workflow, and mode docs revised for the new context model.
• Code review agents: minor refinements to conventions-reviewer and patterns-reviewer.
• .gitignore updated to cover user-preferences, .codex, and dist/.

Removed

• Legacy memory/ directory: runtime memory files (decisions.md, lessons.md, metrics-log.md, research-cache.md, session-state.md) removed from the repo root; replaced by templates under templates/project-info-generation/ that are generated per-project.
• METRICS.md: redundant root-level metrics doc removed — all content was already covered by core/phases/implementation.md, core/rules/memory-handling.md, and core/system/evaluation-flywheel.md.