ai-integration-phaselisted

<purpose> Orchestrate a four-agent pipeline that produces a comprehensive AI-SPEC.md before implementation planning begins. Covers framework selection, implementation guidance, domain context, and evaluation strategy. Steps 3 and 4 (ai-researcher and domain-researcher) are SEQUENTIAL — both write to AI-SPEC.md and running them in parallel causes a last-writer-wins race condition. </purpose> <available_agent_types> Valid MaxVision subagent types (use exact names): - maxvision-framework-selector — Interactive decision matrix for AI/LLM framework selection - maxvision-ai-researcher — Researches chosen framework docs, writes implementation guidance - maxvision-domain-researcher — Researches business domain, surfaces evaluation criteria - maxvision-eval-planner — Designs evaluation strategy, rubrics, guardrails, monitoring plan </available_agent_types> <process> ## 0. Initialize ```bash set -euo pipefail INIT=$(maxvision-sdk query init.plan-phase "${PHASE_ARG:-$ARGUMENTS}") if [[ "$INIT" == @file:* ]]; then INIT=$(cat "${INIT#@file:}"); fi AGENT_SKILLS_SELECTOR=$(maxvision-sdk query agent-skills maxvision-framework-selector) AGENT_SKILLS_AI=$(maxvision-sdk query agent-skills maxvision-ai-researcher) AGENT_SKILLS_DOMAIN=$(maxvision-sdk query agent-skills maxvision-domain-researcher) AGENT_SKILLS_EVAL=$(maxvision-sdk query agent-skills maxvision-eval-planner) ``` Parse JSON for: `phase_dir`, `phase_number`, `phase_name`, `phase_slug`, `padded_phase`, `phase_found`, `has_context