docs(plans): add vision, architecture context, and README requirements to execution guide

- Vision section with full lifecycle and zero-barrier value proposition
- Wave 0 prompt enriched with architectural context (ultraplan as engine,
  pause/resume, MCP phases, marketplace recommendation)
- Session 8 prompt with critical decisions: thin orchestrator architecture,
  ultraplan detection, README value proposition, honest deployment trade-offs

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

.claude/plans/execution-guide.md

@@ -1,5 +1,28 @@
 # Agent Factory — Execution Guide
 
+## Vision
+
+Agent Factory gjør det mulig for alle å lage egne agentløsninger som gjør
+reelle oppgaver — gjennom en guidet, profesjonell prosess fra ide til produksjon.
+
+**Full lifecycle:** Ide → Design → Build → Test → Deploy → Monitor → Improve
+
+**Zero barrier:** Brukeren installerer ktg-plugin-marketplace og får Agent Factory +
+ultra-suiten (ultraplan, ultraresearch, ultraexecute). Ingenting annet trengs —
+ingen servere, ingen databaser, ingen konfigurasjon. Bare Claude Code og en plugin.
+
+**Arkitektur:** Pluginens verdi er domenekunnskap (hva som utgjør et godt agent-system).
+Ultraplan er motoren (hvordan det bygges profesjonelt). Pluginen er en tynn orkestrator
+som kjører et domenespesifikt intervju, mater resultatet til ultraplan, og lar
+ultraexecute implementere planen.
+
+**Deployment er et valg:** Lokal (cron/launchd), VPS (systemd), Docker, eller
+Managed Agents (begrenset — ingen MCP-støtte). Alle targets med full MCP
+er likeverdige. Brukeren velger basert på sine behov.
+
+**Inspirasjon, ikke konkurranse:** OpenClaw og Paperclip er referanseimplementasjoner
+vi lærer av. Agent Factory konkurrerer ikke med dem — vi løser et annet problem.
+
 ## Overview
 
 The ultraplan (`ultraplan-2026-04-11-agent-factory.md`) has 27 steps across 8 sessions
@@ -43,6 +66,13 @@ If not using ultraexecute, use this prompt:
 ```
 Agent Factory Wave 0: Create session blueprints.
 
+VISION (read this first — it shapes everything):
+Agent Factory gjør det mulig for alle å lage egne agentløsninger som gjør
+reelle oppgaver. Full lifecycle: Ide → Design → Build → Test → Deploy → Monitor → Improve.
+Zero barrier: bruker installerer ktg-plugin-marketplace, får Agent Factory + ultra-suiten.
+Ingenting annet trengs. Pluginen er en tynn orkestrator med domenekunnskap — ultraplan
+er motoren. OpenClaw og Paperclip er inspirasjon, ikke konkurrenter.
+
 Context:
 - Plan: .claude/plans/ultraplan-2026-04-11-agent-factory.md
 - Spec: .claude/ultraplan-spec-2026-04-11-agent-factory.md
@@ -79,6 +109,20 @@ For exact template content, use:
   - Source code analysis for implementation details (heartbeat format, budget schema, etc.)
   - Existing codebase files for conventions (frontmatter format, placeholder syntax, hook patterns)
 
+ARCHITECTURAL CONTEXT for blueprints:
+  - Build command (Session 8) should be a thin orchestrator: run domain-specific
+    interview → produce spec → feed to ultraplan → ultraexecute implements.
+    Standalone mode as fallback when ultraplan is not installed.
+  - Build command includes pause/resume mechanism (build-state.json) for when
+    users need to create MCP servers or skills between phases.
+  - Phase 3.5: Skills connection (existing or custom, with pause/resume)
+  - Phase 4.5: MCP server integration (connect existing, guide creation, pause/resume)
+  - Deployment targets: cron/launchd, systemd, Docker (full MCP support).
+    Managed Agents as limited option (no MCP — only simple agents).
+  - README must prominently recommend ktg-plugin-marketplace installation.
+  - README value proposition: "Install one plugin, build complete agent systems
+    that do real work. No servers, no databases, no configuration."
+
 All bash scripts must be bash 3.2 compatible. All templates use {{PLACEHOLDER}} syntax.
 Python3 for JSON/YAML/date parsing in scripts.
 
@@ -308,6 +352,23 @@ Execute steps 8, 26, 27:
     - Resume mechanism: --resume flag reads build-state.json
 27. Update plugin.json to v1.0.0, rewrite CLAUDE.md and README.md for full Agent Factory
 
+CRITICAL ARCHITECTURAL DECISIONS for Step 26 (build command):
+- Build command should be a THIN ORCHESTRATOR: run domain-specific interview,
+  produce spec, feed to ultraplan-local if available, fall back to standalone if not.
+- Detect ultraplan availability: check if /ultraplan-local skill exists.
+  If yes: offer "Professional mode (recommended): uses ultraplan for deep planning"
+  If no: run standalone build workflow (current monolithic approach)
+- Build must support pause/resume: build-state.json persists all state between sessions
+
+CRITICAL for Step 27 (README):
+- README must prominently recommend ktg-plugin-marketplace:
+  "## Prerequisites\n\nInstall the plugin marketplace for the full experience:\n\n```\n/install ktg-plugin-marketplace\n```\n\nThis gives you Agent Factory + the ultra-suite (ultraplan, ultraresearch,\nultraexecute) for professional-grade agent system development."
+- Value proposition: "Install one plugin, build complete agent systems that do
+  real work. Full lifecycle from idea to production. No servers, no databases,
+  no configuration — just Claude Code."
+- Deployment targets section with honest trade-offs (Managed Agents = limited, no MCP)
+- Credit OpenClaw + Paperclip as inspiration, not competitors
+
 Steps 8 and 26 both modify commands/build.md — execute them SEQUENTIALLY.
 Step 27 is the final commit: "feat: Agent Factory v1.0.0"