Agent Orchestration Showcase - Setup Playbook & Cheat Sheet

Last Updated: 2026-03-24 Status: Research & Planning Complete. Ready for Implementation. Project Owner: Dmitri Zasage

What Is This Project?
Current Status
File Map (Where Everything Lives)
Architecture Overview
Tool Stack
n8n Setup Guide
Content Pipeline
Brand Quick Reference
What’s Left To Do
How To Resume In a New Session
Key Documentation Links
Research Findings Summary
Troubleshooting & Known Issues

1. What Is This Project? {#what-is-this-project}

Solanasis is building 3 showcase AI agent orchestration demos, one for each primary ICP:

Demo	Target ICP	Agents	Key Hook
AI Legal Ops Agent	Estate planning attorneys (5-50 person firms)	6-agent pipeline	Contract review in 4 min vs 4 hrs
Foundation Ops Agent	Private foundations ( $5 M -$ 50M assets, 1-10 staff)	7-agent pipeline	20 grant applications reviewed in 2 hrs
SMB Health Check	SMBs (11-200 employees, no CISO/CIO)	6-agent pipeline	Automated operational resilience assessment

Why: Need tangible, working demos (not just pitch decks) that prove Solanasis can set up AI-powered operational systems. Each demo becomes:

A case study blog post
A 5-post LinkedIn series
A 3-5 minute video walkthrough
A branded PDF one-pager for sales
An outreach email template

Cost: $0 beyond existing Claude Code Max subscription. n8n is free (self-hosted). Typst is free (open-source).

2. Current Status {#current-status}

What’s Been Done (2026-03-24)

Master plan written with full requirements, architecture, market intelligence
Phase 1 (Attorney Demo) sub-plan written with 10 implementation steps
Phase 2 (Foundation Demo) sub-plan written with 9 implementation steps
5 parallel research agents completed:
- n8n programmatic setup and automation (API, CLI, JSON import)
- Solanasis existing content patterns (brand kit, blog system, voice profile)
- Canva MCP capabilities (generate, export, limitations)
- Astro blog publishing structure (frontmatter, collections, deployment)
- AI-native agency best practices (demo strategy, content pipelines, PDF tools)
Senior reviewer validated plan; 6 issues found and fixed
All plan files updated with research findings
Brand colors corrected across all plans
Implementation steps reconciled with research recommendations

What’s NOT Done Yet

n8n not installed (Docker Compose config not yet created)
Typst not installed
No synthetic demo data created
No n8n workflows built
No content written (blog, LinkedIn, outreach)
No one-pager template created
No video recorded
SMB demo (Phase 3) is deferred

3. File Map {#file-map}

Plan Files (The Source of Truth)

File	What It Is	Path
Master Plan	Strategy, architecture, market intel, research findings	`.claude-plans/deep-plan-agent-orchestration-showcase-2026-03-24.md`
Phase 1 Plan	Attorney demo: 10 implementation steps	`.claude-plans/deep-plan-agent-orchestration-showcase-phase-1-attorney-2026-03-24.md`
Phase 2 Plan	Foundation demo: 9 implementation steps	`.claude-plans/deep-plan-agent-orchestration-showcase-phase-2-foundation-2026-03-24.md`
This Playbook	You’re reading it	`solanasis-docs/playbooks/agent-orchestration-showcase-playbook.md`

Reference Playbooks (Feed Into Demo Content)

File	Relevance
`playbooks/Estate_Planning_Attorney_Smartcut_Playbook.md`	Attorney ICP pain points, pitch scripts, service mapping
`playbooks/Foundation_and_Planned_Giving_Prospecting_Playbook_2026-03-16.md`	Foundation economics, compliance needs, decision-making process

Brand & Content References

File	What It Is
`brand-style/solanasis-brand-kit.md`	Canonical brand identity (v2.0, March 2026) — colors, fonts, logos, spacing
`brand-style/Brand_Style_Guide.md`	Extended guide with 14 sections
`brand-style/Color Palette.md`	Quick color reference with CSS tokens
`solanasis-voice-profile.md`	Voice/tone rules for all content
`content-creation/LinkedIn_Posting_Cheat_Sheet.md`	60 ready-to-post LinkedIn options
`content-creation/Image_Style_Guide_v2.md`	Pencil sketch image generation style
`content-creation/Solanasis_Image_Generation_SOP.md`	Full image generation pipeline

Output Directories (Where New Files Go)

Content Type	Save To	Exists?
Blog posts	`solanasis-site/src/content/blog/`	Yes
Blog images	`solanasis-site/public/images/stock/`	Yes
LinkedIn posts	`solanasis-docs/content-creation/linkedin-posts/`	Create
One-pager PDFs	`solanasis-docs/marketing_assets/sales-collateral/`	Create `sales-collateral/`
Demo data (attorney)	`solanasis-docs/reference/demo-data/attorney-demo/`	Create
Demo data (foundation)	`solanasis-docs/reference/demo-data/foundation-demo/`	Create
Architecture diagrams	`solanasis-docs/content-creation/`	Yes
Outreach templates	`solanasis-docs/daily-outreach/`	Yes
n8n setup files	`solanasis-scripts/n8n-setup/`	Create

Pro tip: Directories marked “Create” don’t exist yet. They’ll be created during Phase 1 execution. Don’t create empty directories in advance; let them be created when the first file goes in.

4. Architecture Overview {#architecture-overview}

The Stack (All Free)

+-----------------------------------------------------------+
|                 SOLANASIS AI OPS LAYER                     |
+-----------------------------------------------------------+
|                                                           |
|  CLAUDE CODE          n8n                 PYTHON          |
|  (Brain)              (Visual Workflow)   (Glue)          |
|                                                           |
|  - Skills             - Triggers          - PDF parse     |
|  - Subagents          - Visual UI         - API calls     |
|  - MCP servers        - Webhooks          - Data xform    |
|  - Hooks              - 400+ apps         - Reports       |
|                                                           |
|  Connected via:       Connected via:      Connected via:  |
|  - ClickUp MCP       - Anthropic API     - Typst CLI     |
|  - Gmail MCP         - ClickUp node      - n8n Execute   |
|  - Calendar MCP      - Gmail node          Command node  |
|  - Canva MCP         - Calendar node                     |
|  - Fireflies MCP     - HTTP Request                      |
+-----------------------------------------------------------+

Attorney Demo: 6-Agent Pipeline

Document Upload --> [1] Document Parser (Haiku)
                --> [2] Clause Analyzer (Sonnet) -- GREEN/YELLOW/RED risk flags
                --> [3] Jurisdiction & Compliance (Sonnet)
                --> [4] Senior Reviewer (Opus) -- quality score 0-100
                --> [5] Report Generator (Sonnet) -- executive summary + redlines
                --> [6] Action Router (Haiku) -- ClickUp tasks + Gmail draft + Calendar event

Cost per run: ~$1.27 (mostly the Opus reviewer)

Foundation Demo: 7-Agent Pipeline

Grant Application --> [1] Application Parser (Haiku)
                  --> [2] Due Diligence Researcher (Sonnet + web search)
                  --> [3] Mission Alignment Scorer (Sonnet) -- 0-100 score
                  --> [4] Financial Analyst (Sonnet) -- overhead ratio, sustainability
                  --> [5] Compliance Checker (Sonnet) -- expenditure responsibility, self-dealing
                  --> [6] Board Summary Generator (Opus) -- APPROVE/DECLINE/MORE INFO
                  --> [7] Action Router (Haiku) -- ClickUp card + Calendar + Gmail

Cost per application: ~$1.34

n8n Implementation Pattern

Each agent is its own sub-workflow in n8n. An orchestrator workflow calls them via the Call n8n Workflow Tool node. This means:

Each agent has its own system prompt, Claude model, and tools
Shared context passes via Postgres Chat Memory (common sessionId)
Workflows are buildable as JSON and importable in one shot
n8n AI-specific connection types: ai_languageModel, ai_tool, ai_memory

5. Tool Stack {#tool-stack}

Tool	Role	Status	Install
Docker Desktop	Container runtime for n8n	Installed (v28.3.3)	Already done
Python	Glue scripts, data parsing	Installed (3.12.0)	Already done
n8n (Docker Compose)	Visual workflow automation	NOT INSTALLED	See Section 6
Typst	PDF one-pager generation	NOT INSTALLED	`winget install typst`
Claude Code	AI orchestration engine	Installed (Max plan)	Already done
OBS Studio	Screen recording for demo videos	Check if installed	Free download
Mermaid.js	Architecture diagrams	Built into markdown	No install needed

Optional Tools (Deferred)

Tool	Role	When to Install
n8n-MCP (`czlonkowski/n8n-mcp`)	Let Claude Code manage n8n workflows directly	If JSON workflow building proves cumbersome
Taplio or Buffer	LinkedIn post scheduling	When ready to publish LinkedIn series
LangChain/CrewAI	Alternative agent frameworks for comparison blog	Phase 4 (deferred)

6. n8n Setup Guide {#n8n-setup-guide}

Step-by-Step Installation

1. Create the setup directory:

solanasis-scripts/n8n-setup/
  docker-compose.yml
  .env              (gitignored!)
  .gitignore        (contains: .env)

2. Generate secure credentials:

# Encryption key for n8n credential storage
openssl rand -hex 24
# Result goes into .env as N8N_ENCRYPTION_KEY
 
# Postgres password
openssl rand -hex 16
# Result goes into .env as POSTGRES_PASSWORD

3. The .env file should contain:

N8N_ENCRYPTION_KEY=<generated-hex>
POSTGRES_PASSWORD=<generated-hex>

4. Docker Compose config (full version in master plan, key settings):

PostgreSQL 16 for persistent storage
n8n latest image
Port 5678 exposed
Timezone: America/Denver
Telemetry disabled
Execution data auto-pruned after 168 hours (7 days)
Volumes: n8n-data, n8n-files, n8n-postgres-data

5. Start it:

cd solanasis-scripts/n8n-setup/
docker compose up -d

6. First-time setup (Chrome extension, ~2 minutes):

Navigate to http://localhost:5678
Create owner account (email + password)
Save credentials to .env or password manager
Go to Settings > n8n API > Create API Key
Save API key to .env

7. Create credentials programmatically:

# Anthropic API key
curl -X POST http://localhost:5678/api/v1/credentials \
  -H "X-N8N-API-KEY: your-n8n-api-key" \
  -H "Content-Type: application/json" \
  -d '{"name": "Anthropic API Key", "type": "anthropicApi", "data": {"apiKey": "sk-ant-..."}}'

8. OAuth2 for Gmail/Calendar:

Must be done via Chrome extension (interactive browser auth)
~2 minutes per service

9. Import workflows:

# Mount workflow JSONs into container, then import
docker exec -u node -it n8n n8n import:workflow --input=/files/all-workflows.json

10. Activate workflows:

curl -X POST http://localhost:5678/api/v1/workflows/{id}/activate \
  -H "X-N8N-API-KEY: your-n8n-api-key"

n8n Key Concepts

AI Agent node: @n8n/n8n-nodes-langchain.agent — supports Tools Agent, ReAct, Plan-and-Execute
Anthropic Chat Model: @n8n/n8n-nodes-langchain.lmChatAnthropic — connects via ai_languageModel
Sub-workflows: Each agent lives in its own workflow; orchestrator calls them via toolWorkflow node
Memory: Postgres Chat Memory with shared sessionId for context between agents
Export: docker exec -u node -it n8n n8n export:workflow --all --output=/files/backup.json

n8n REST API Quick Reference

Action	Method	Endpoint
List workflows	GET	`/api/v1/workflows`
Create workflow	POST	`/api/v1/workflows`
Activate	POST	`/api/v1/workflows/:id/activate`
Execute	POST	`/api/v1/workflows/:id/execute`
List credentials	GET	`/api/v1/credentials`
Create credential	POST	`/api/v1/credentials`
Get credential schema	GET	`/api/v1/credentials/schema/:type`

All endpoints require header: X-N8N-API-KEY: your-key

Cleanup / Reset

# Stop n8n
docker compose down
 
# Full reset (destroys all data)
docker compose down -v
 
# Restart
docker compose up -d

7. Content Pipeline {#content-pipeline}

Blog Post Publishing (Fully Automated)

Write .md file with frontmatter
    |
    v
Generate sketch-style image (1200x800 JPG + WebP)
    |
    v
Place .md in solanasis-site/src/content/blog/
Place images in solanasis-site/public/images/stock/
    |
    v
Update BLOG-QUEUE.md
    |
    v
Git commit + push to main
    |
    v
GitHub Actions daily cron (6am MT) rebuilds + deploys to Cloudflare Pages
    |
    v
Post is live at solanasis.com/blog/{slug}

Required frontmatter:

---
title: "How We Built a 6-Agent AI Legal Ops System in One Afternoon"
description: "A working demo of multi-agent AI contract review for law firms."
date: 2026-04-XX        # Target a Thursday; auto-publishes when date <= today
author: "Solanasis Team"
tags: ["ai-agents", "automation", "legal-tech", "n8n"]
pillar: "ai-automation"  # Must be one of 8 valid values
image: "/images/stock/blog-ai-legal-ops-agent.jpg"
imageAlt: "Pencil sketch of AI agents reviewing legal documents"
readingTime: 7           # ~200 words per minute
draft: false
---

Valid pillars: cybersecurity, disaster-recovery, data-systems, crm-operations, ai-automation, operational-resilience, founder-journey, partnerships

Image specs:

1200x800px (3:2 aspect ratio)
JPG (quality 85) + WebP (quality 80)
Pencil sketch style via Nano Banana (gemini-2.5-flash-image)
Filename: blog-{slug}.jpg and blog-{slug}.webp
Scripts: solanasis-scripts/generate_all_sketch_images.py

Create content (markdown or direct Typst)
    |
    v
Apply Solanasis-branded Typst template
(Ink Navy #020532, Copper #C47A3D, Playfair Display, Montserrat, logo)
    |
    v
Compile: typst compile one-pager.typ one-pager.pdf
    |
    v
Save to solanasis-docs/marketing_assets/sales-collateral/

Typst CLI usage:

# Basic compile
typst compile template.typ output.pdf
 
# With variables (for personalized one-pagers)
typst compile --input service="Legal Ops" --input icp="Attorney" template.typ output.pdf

Fallback: Canva MCP generate-design + export-design for higher visual polish (budget 10-15 min manual refinement in Canva editor).

LinkedIn Post Series (5 Posts Per Demo)

Three posting styles (pick based on energy):

Style	Vibe	Best For
A: Casual Teacher	Mid-thought opener, teaches as you go	Explaining how something works
B: Sharp Take	Bold opinion, backed up	Challenging conventional wisdom
C: Story Drop	Personal experience, pivots to lesson	Case studies, war stories

Voice rules:

First person (“I”)
1-2 sentences per paragraph, 6-14 lines per post
No em dashes (use semicolons, commas, parentheses)
No engagement-bait questions, no CTAs in posts
60% professional / 40% raw
Banned: “game-changer”, “next level”, “leverage”, “synergy”

Save to: solanasis-docs/content-creation/linkedin-posts/

Content Atomization (The Flywheel)

Each demo produces all of these assets:

Working Demo
    |
    +-> Blog post (1,200-1,800 words, pillar: ai-automation)
    +-> LinkedIn series (5 posts, 3 style options each)
    +-> Video walkthrough (3-5 min, Dmitri records with OBS)
    +-> Architecture diagram (Mermaid for blog, Canva for LinkedIn)
    +-> One-pager PDF (Typst, branded)
    +-> Outreach email template

8. Brand Quick Reference {#brand-quick-reference}

Colors (Canonical — from solanasis-brand-kit.md v2.0)

Token	Name	Hex	RGB	Usage
Primary	Ink Navy	`#020532`	2, 5, 50	Nav, hero, headers, dark sections
Secondary	Deep Navy	`#091652`	9, 22, 82	Links, icons, card borders
Accent	Copper	`#C47A3D`	196, 122, 61	CTA buttons, badges, highlights
Background	Parchment	`#FEF9F1`	254, 249, 241	Page base, light sections
Text	Charcoal	`#111827`	17, 24, 39	Body copy, labels

Typography

Role	Font	Weights
Display/Headings	Playfair Display	400, 600, 700
Body/UI	Montserrat	300-700
Accent/Subtitles	Libre Baskerville	400, 400i, 700
Code/Data	JetBrains Mono	400

Logos

Location: solanasis-docs/logo/

File	Use For
`Solanasis - Logo - Squared.png`	Social media, favicons
`Solanasis - Logo Horizontal.png`	Headers, decks (primary)
`Solanasis - Logo Icon.png`	Small spaces, app icons
`Solanasis logo for dark background transparent.png`	Dark overlays

Brand Essence

“Grounded. Sharp. Human.” — Professional without corporate stuffiness.

9. What’s Left To Do {#whats-left-to-do}

Phase 1: Attorney Demo (Estimated 7-9 hours with automation)

Track A: Infrastructure (can run in parallel with Track B)

Track B: Content Production (can run in parallel with Track A)

Create synthetic demo data (3 sample legal documents: NDA, engagement letter, vendor contract)
Create Mermaid architecture diagram for blog
Create Solanasis-branded Typst one-pager template
Draft blog post with proper Astro frontmatter
Draft 5 LinkedIn posts
Draft outreach email template
Generate blog header image (Nano Banana sketch style)
Compile one-pager PDF via Typst

Track C: Demo Polish (after A + B complete)

Full dress rehearsal of pipeline
Tune system prompts based on output quality
Target: under 5 minutes per document, quality “good enough for a managing partner”
Dmitri records video with OBS (3-5 min)
Publish blog post (git push to main)
Schedule LinkedIn posts

Phase 2: Foundation Demo (Estimated 6-8 hours; n8n carries forward)

Create 5 synthetic grant applications
Create foundation mission profile document
Build 7 agent sub-workflow JSONs + 1 orchestrator
Import and test
Blog post + LinkedIn series + one-pager + outreach email
Video recording
Contact at least 3 foundation prospects

Phase 3: SMB Demo (Deferred)

Only if Phases 1-2 show traction with prospects.

Phase 4: LangChain Comparison Blog (Deferred)

SEO content targeting technical readers. 4-6 hours.

10. How To Resume In a New Session {#how-to-resume}

Quick Start Prompt

Copy-paste this into a new Claude Code session:

I'm picking up the Agent Orchestration Showcase project. The playbook is at
solanasis-docs/playbooks/agent-orchestration-showcase-playbook.md and the
detailed plans are in .claude-plans/deep-plan-agent-orchestration-showcase-*.md.
Read the playbook first, then let's start executing Phase 1.

What Claude Code Needs to Know

Plan files are in .claude-plans/ — three files (master + phase 1 + phase 2)
Research findings are appended to the master plan — start reading from the “RESEARCH FINDINGS” section (~line 500)
Phase 1 has research-informed implementation notes appended after the risks table
Memory files exist at ~/.claude/projects/C---my--solanasis/memory/:
- project_agent_orchestration_showcase.md — project context
- feedback_automate_everything.md — automation-first approach

Key Decisions Already Made

Decision	Choice	Why
n8n setup method	Docker Compose with PostgreSQL	Persistent storage, production-ready
PDF one-pager tool	Typst (primary), Canva MCP (fallback)	Typst is fully automated CLI; Canva needs 10-15 min polish
Blog pillar	`ai-automation`	Existing pillar, perfect fit
Agent architecture	Sub-workflows with Postgres Chat Memory	Each agent isolated, shared context via sessionId
Demo style	Sequential pipeline (left-to-right)	Clearest for non-technical audiences
n8n visual style	Color-coded nodes, sticky notes, renamed nodes	Best practices from AI agency research
n8n-MCP tool	Deferred/optional	Only install if JSON workflow building is cumbersome

11. Key Documentation Links {#documentation-links}

n8n

Resource	URL
n8n Docker Installation	https://docs.n8n.io/hosting/installation/docker/
n8n REST API Reference	https://docs.n8n.io/api/api-reference/
n8n CLI Commands	https://docs.n8n.io/hosting/cli-commands/
n8n Environment Variables	https://docs.n8n.io/hosting/configuration/environment-variables/
n8n AI Agent Node	https://docs.n8n.io/integrations/builtin/cluster-nodes/root-nodes/n8n-nodes-langchain.agent/
n8n Anthropic Chat Model Node	https://docs.n8n.io/integrations/builtin/cluster-nodes/sub-nodes/n8n-nodes-langchain.lmchatanthropic/
n8n Call Workflow Tool	https://docs.n8n.io/integrations/builtin/cluster-nodes/sub-nodes/n8n-nodes-langchain.toolworkflow/
n8n Postgres Chat Memory	https://docs.n8n.io/integrations/builtin/cluster-nodes/sub-nodes/n8n-nodes-langchain.memorypostgreschat/
n8n Export/Import Workflows	https://docs.n8n.io/workflows/export-import/
n8n Multi-Agent Systems Blog	https://blog.n8n.io/multi-agent-systems/
n8n AI Workflow Templates	https://n8n.io/workflows/categories/ai/
n8n Workflow JSON Guide (GitHub)	https://github.com/nordeim/n8n-Workflow/blob/main/authoritative_n8n_workflow_json_guide.md
n8n-MCP Tool (GitHub)	https://github.com/czlonkowski/n8n-mcp
n8n Security Best Practices	https://www.soraia.io/blog/n8n-security-best-practices-protect-your-data-and-workflows

Typst (PDF Generation)

Resource	URL
Typst Official Docs	https://typst.app/docs/
Typst CLI Reference	https://typst.app/docs/reference/cli/
Typst Automated Generation Blog	https://typst.app/blog/2025/automated-generation/
Pandoc + Typst Guide	https://neilzone.co.uk/2025/01/using-pandoc-and-typst-to-convert-markdown-into-custom-formatted-pdfs-with-a-sample-template/
Typst GitHub Releases	https://github.com/typst/typst/releases

Astro & Cloudflare (Blog)

Resource	URL
Astro Content Collections	https://docs.astro.build/en/guides/content-collections/
Astro Cloudflare Adapter	https://docs.astro.build/en/guides/integrations-guide/cloudflare/
Cloudflare Pages Docs	https://developers.cloudflare.com/pages/

Canva MCP

Resource	URL
Canva MCP Usage Guide	https://www.canva.com/help/mcp-canva-usage/
Canva MCP Server Setup	https://www.canva.dev/docs/connect/canva-mcp-server-setup/
Canva Dev API Reference	https://www.canva.dev/docs/connect/api-reference/
Canva Autofill Guide	https://www.canva.dev/docs/connect/autofill-guide/

Market Intelligence (Referenced in Plan)

Resource	URL
Legal AI Access at 83% (2026)	https://www.artificiallawyer.com/2026/03/23/legal-ai-access-at-83-but-trust-issues-remain/
AI Law Firms to Watch	https://www.lupl.com/blog/10-ai-law-firms-to-watch-in-2026/
AI Tools for Nonprofits	https://bloomerang.com/blog/ai-tools-for-nonprofits/
AI Workflow Automation Agency Guide	https://www.heyreach.io/blog/ai-workflow-automation-agency
Relevance AI ($24M Series B)	https://techcrunch.com/2025/05/06/relevance-ai-raises-24m-series-b-to-help-anyone-build-teams-of-ai-agents/

Demo & Content Strategy

Resource	URL
n8n Design Pattern with Color-Coding	https://n8n.io/workflows/9500-standardized-workflow-design-pattern-with-color-coding-system-for-teams/
20 n8n Tricks for Pro Builders	https://www.aifire.co/p/20-n8n-tricks-for-next-level-automation-building-part-1
AI Demo Tools for B2B Sales	https://goconsensus.com/blog/ai-demo-tools-for-sales
Content Pipeline Automation	https://www.trysight.ai/blog/automated-blog-content-pipeline
Building Content Pipelines (Markdown to PDF)	https://dasroot.net/posts/2026/03/building-content-pipelines-markdown-pdf-epub-html/
LinkedIn Scheduling Tools Compared	https://www.supergrow.ai/blog/linkedin-scheduling-tools

12. Research Findings Summary {#research-summary}

n8n Automation (Research Agent 1)

90%+ automatable. Docker Compose + REST API + CLI covers nearly everything.
Owner account is the one friction point — no env var to skip it. Use Chrome extension (~30 sec).
Workflow JSON import is robust. Build as JSON, import via CLI, activate via API.
Sub-workflow pattern is the right architecture for multi-agent orchestration.
Postgres Chat Memory for shared context between agents (common sessionId).
Known bug: Anthropic credential validation sometimes fails with valid keys (GitHub #15646, #26300). Retry works.
8,856+ community templates available, including AI agent examples.
Credential type for Anthropic: anthropicApi (field: apiKey).

Solanasis Content Patterns (Research Agent 2)

Brand system is complete and professional-grade. Colors, fonts, logos, spacing tokens all documented.
Blog publishing is fully automated — Astro + Cloudflare + GitHub Actions daily cron.
No one-pager template exists yet, but all components (brand kit, design tokens) are ready.
Voice profile is defined: no em dashes, first person, 3 LinkedIn styles, specific banned words.
60 LinkedIn posts are already drafted and ready in the Cheat Sheet.
Image generation pipeline exists — pencil sketch style via Nano Banana.
generate-pitch-deck.py exists as a pattern for programmatic document generation.
Marketing assets dir is marketing_assets/ (underscore) not marketing-assets/ (hyphen).

Canva MCP (Research Agent 3)

80% automatable, 20% manual polish for client-facing quality.
generate-design creates multiple candidates from natural language prompt.
export-design supports PDF, PNG, JPG, SVG, PPTX.
Editing pipeline is limited — only text/media replacement. No font, color, or layout changes.
Brand template autofill is Enterprise-only.
list-brand-kits can read brand data on any paid plan.
upload-asset-from-url works for pushing logos/diagrams into Canva.
Recommendation: Typst for fully automated; Canva MCP for visual polish when needed.

Astro Blog (Research Agent 4)

Content collection in src/content/blog/ with strict TypeScript schema validation.
8 valid pillar enum values — typos break the build.
Posts auto-publish when date <= today during daily 6am MT rebuild.
BLOG-QUEUE.md is the single source of truth for scheduling.
Publishing cadence: Thursdays.
Image format: blog-{slug}.jpg + blog-{slug}.webp at 1200x800.
generate-pdfs.mjs script exists for PDF generation.
Git push to main triggers deployment.

AI-Native Agency Patterns (Research Agent 5)

“Self-demonstration” is highest-converting — use your own automation to prospect before the meeting.
Sequential pipeline is the clearest demo pattern for non-technical audiences.
Content atomization: 1 case study = 8 LinkedIn posts + blog + PDF + outreach.
n8n visual best practices: color-code nodes, sticky notes as headers, rename every node, left-to-right flow.
Typst is the recommended PDF tool — millisecond compilation, CLI-friendly, JSON data input.
OBS Studio for recording; structure demos in 90-second segments.
MCP-powered workflows command 2-4x premium over basic automation pricing.
Pricing benchmark: AI agent pilots = $10 K -$ 40K market rate; Solanasis can undercut at $3 K -$ 5K as door-opener.

13. Troubleshooting & Known Issues {#troubleshooting}

Issue	Cause	Fix
n8n Anthropic credential “resource not found”	Known n8n bug (#15646, #26300)	Retry credential creation; test with curl first
Docker volume permission errors on Windows	WSL2 + Docker mount mismatch	Use named volumes (not bind mounts) for n8n data
Blog build fails after adding post	Pillar enum typo or missing required field	Check frontmatter against the 8 valid pillars and schema
n8n workflow import overwrites existing	Imported IDs match existing workflow IDs	Strip/change IDs in JSON before importing
Canva MCP “Missing scopes” error	OAuth token needs refresh	Disconnect and reconnect the Canva connector
Exported n8n credentials have no secrets	Default export doesn’t include secrets	Use `--decrypted` flag: `n8n export:credentials --all --decrypted`
OAuth2 credentials can’t be created via API	Requires interactive browser flow	Use Chrome extension for Gmail/Calendar OAuth
Typst font not found	System font not installed	Install Playfair Display + Montserrat, or use `--font-path` flag

Revision History

Date	Change	By
2026-03-24	Initial plan creation (master + phase 1 + phase 2)	Claude Code (Opus 4.6)
2026-03-24	5 research agents completed; findings integrated into plans	Claude Code (Opus 4.6)
2026-03-24	Senior review: 6 issues found and fixed	Claude Code (senior-reviewer)
2026-03-24	This playbook/cheat sheet created	Claude Code (Opus 4.6)

Solanasis Docs

Explorer

agent-orchestration-showcase-playbook