davidcole.cloud — Agent Swarm Graph

Master map of the David Cole / davidcole.cloud agent layer — how named agents connect, when they fire, and how they tie into the existing Gnomad Studio swarm (ZED, NEO, NIOBE, ORACLE, MORPHEOUS, etc.).

  • davidcole-swarm
  • reference

davidcole.cloud — Agent Swarm Graph

Master map of the David Cole / davidcole.cloud agent layer — how named agents connect, when they fire, and how they tie into the existing Gnomad Studio swarm (ZED, NEO, NIOBE, ORACLE, MORPHEOUS, etc.).


Agent Roster (Names)

NameFrameworkDomainOne-line role
CHORUSPraisonAI / CrewAIPortfolio & Inbound MarketingMulti-agent crew: local SEO research + copy → case studies
HAWKCustom intelligence agentPortfolio & Inbound MarketingScrapes/analyzes regional competitors for gaps and keywords
NEXUSMCPTechnical Sandbox & BlogBridges IDEs to Linux filesystem/terminal via MCP
FORGEOpenHands / TaskWeaverTechnical Sandbox & BlogIsolated code/command verification before publish
AEONAeon + GitHub ActionsInfrastructure & Cloud AutomationAutonomous CI: build, quality, SEO audits on every push
GUIDEPhidata / AgnoInfrastructure & Cloud AutomationPortfolio concierge with memory + KB for client inquiries
QUILLCustom chroniclerCross-cuttingRecords sessions → drafts /notes blog posts

Existing Gnomad directors still orchestrate: ZED (PM), ORACLE (research), MORPHEOUS (MCP/AAIO strategy), NEO, NIOBE, OPTIMUS, KEYMAKER.


System Graph (who talks to whom)

flowchart TB
  subgraph userLayer [You — David Cole]
    DAVID[David]
  end

  subgraph gnomadCore [Gnomad Swarm Core]
    ZED[ZED — PM Orchestrator]
    ORACLE[ORACLE — Research Director]
    MORPHEOUS[MORPHEOUS — AI Architect]
    NEO[NEO — Engineering]
    OPTIMUS[OPTIMUS — SEO Growth]
    KEYMAKER[KEYMAKER — Vault Ops]
  end

  subgraph portfolio [Portfolio and Inbound]
    HAWK[HAWK — Competitor Intel]
    CHORUS[CHORUS — Case Study Crew]
  end

  subgraph technical [Technical Sandbox and Blog]
    NEXUS[NEXUS — MCP Bridge]
    FORGE[FORGE — Code Sandbox]
  end

  subgraph infra [Infrastructure]
    AEON[AEON — CI Automation]
    GUIDE[GUIDE — Portfolio Concierge]
  end

  subgraph memory [Memory and Publishing]
    QUILL[QUILL — Chronicler]
    SITE[davidcole.cloud /notes and /projects]
  end

  DAVID --> ZED
  ZED --> CHORUS
  ZED --> NEXUS
  ZED --> AEON
  ZED --> QUILL

  ORACLE --> HAWK
  HAWK --> CHORUS
  HAWK --> OPTIMUS
  HAWK --> QUILL

  CHORUS --> SITE
  CHORUS --> QUILL
  CHORUS --> FORGE

  MORPHEOUS --> NEXUS
  NEXUS --> FORGE
  NEXUS --> KEYMAKER
  NEXUS --> QUILL

  FORGE --> QUILL
  FORGE --> SITE

  AEON --> SITE
  AEON --> QUILL
  AEON --> ZED
  KEYMAKER --> AEON

  GUIDE --> SITE
  GUIDE --> QUILL
  GUIDE --> CHORUS
  ATLAS[ATLAS — Strategy] --> GUIDE
  ATLAS --> CHORUS

  QUILL --> SITE
  OPTIMUS --> CHORUS
  OPTIMUS --> AEON
  NEO --> NEXUS
  NEO --> AEON

Invocation Map (when each agent runs)

flowchart LR
  subgraph triggers [Triggers]
    T1[Client project delivered]
    T2[New vertical or region]
    T3[IDE needs local tools]
    T4[Blog draft with code]
    T5[git push]
    T6[Visitor on site]
    T7[End of work session]
  end

  T1 --> CHORUS
  T2 --> HAWK
  HAWK --> CHORUS
  T3 --> NEXUS
  T4 --> FORGE
  T5 --> AEON
  T6 --> GUIDE
  T7 --> QUILL

  CHORUS --> QUILL
  NEXUS --> QUILL
  FORGE --> QUILL
  AEON --> QUILL
  GUIDE --> QUILL
  QUILL --> SITE_OUT[Publish to /notes]
  FORGE --> SITE_OUT
  CHORUS --> SITE_OUT
  AEON --> SITE_OUT

Trigger cheat sheet

TriggerPrimary agentWhat happens
You say “new case study for [client]”CHORUSScout researches local SEO; Scribe drafts MDX → /projects
“Who competes in [city]?”HAWKCompetitor matrix → feeds CHORUS + OPTIMUS + QUILL
“Connect Cursor to my homelab”NEXUSMCP server + docs; MORPHEOUS/KEYMAKER review security
Blog post includes shell/PythonFORGERuns in sandbox; pass/fail before publish
git push to site repoAEONBuild, SEO audit, quality gates (unattended)
Visitor asks about servicesGUIDEKB + memory answer; logs gaps for QUILL
“Wrap up today” / session endQUILLSession log + optional draft note (draft: true)

Workflow Lanes (how work flows together)

Lane 1 — Portfolio & inbound marketing

HAWK (market gaps)
  → CHORUS (case study MDX)
  → FORGE (if code in study)
  → OPTIMUS (SEO pass)
  → AEON (build + audit on push)
  → live /projects/[slug]
  → QUILL (narrative: "how we built this")

Invoke CHORUS when: you have a deliverable worth showcasing.
Invoke HAWK first when: you don’t know which vertical or angle to target.

Lane 2 — Technical sandbox & blog

NEXUS (MCP or homelab work)
  → FORGE (verify commands)
  → QUILL (draft /notes MDX)
  → you approve (draft: false)
  → AEON → live /notes/[slug]

Yes — you write blogs about what you work on. QUILL turns real sessions (MCP setups, CI pipelines, agent orchestration) into /notes under:

  • Agentic Workflows — CHORUS, HAWK, swarm patterns
  • Linux/Homelab — NEXUS, FORGE, Nobara tooling
  • DIY Builds — AEON, GUIDE, site features

Lane 3 — Infrastructure & always-on

git push → AEON (automatic, no prompt needed)
Visitor chat → GUIDE (automatic when embedded)
Weekly → QUILL digest (manual or scheduled)

Lane 4 — Chronicler (QUILL) — the glue

QUILL is not a last step only. It listens across the swarm:

UpstreamQUILL produces
CHORUS“How we positioned [industry] case study”
HAWK“Regional competitor landscape Q1”
NEXUS“Building an MCP filesystem bridge on Nobara”
FORGE“Commands that failed vs. worked”
AEON“What CI caught before deploy”
GUIDE“Top 5 prospect questions this month”

Internal logs: .agents/logs/YYYY-MM-DD-session.md
Public face: Website-Stuff/src/content/notes/*.mdx


Integration with Gnomad Swarm (existing agents)

David Cole agentGnomad counterpartRelationship
CHORUSATLAS, ECHO, OPTIMUSATLAS sets angle; ECHO voice; OPTIMUS SEO
HAWKORACLEHAWK is tactical scrape; ORACLE is strategic dossier
NEXUSMORPHEOUS, KEYMAKERMORPHEOUS owns MCP strategy; KEYMAKER owns secrets
FORGENEO, CYPHEREngineering verification before content ships
AEONZED, KEYMAKER, SENTINELZED owns deploy gate; SENTINEL SEO rules in CI
GUIDEATLAS, OPTIMUSMessaging + conversion; embedded on NEO’s Astro stack
QUILLZEDZED’s task.md / walkthrough → QUILL formalizes

Entry point remains ZED. You don’t invoke seven agents manually — you tell ZED the goal; ZED assigns CHORUS, HAWK, NEXUS, etc. per SBDT checklist.


Example: one week on davidcole.cloud

sequenceDiagram
  participant D as David
  participant Z as ZED
  participant H as HAWK
  participant C as CHORUS
  participant N as NEXUS
  participant F as FORGE
  participant Q as QUILL
  participant A as AEON
  participant S as Site

  D->>Z: Ship CRM case study for local dentists
  Z->>H: Gap analysis — dental web design, Tulsa
  H-->>C: Keywords + competitor angles
  C-->>S: Draft projects/dental-crm.mdx
  C->>Q: Session + publish brief
  D->>N: MCP server for local file search
  N->>F: Verify setup scripts
  F-->>Q: Approved snippets
  Q-->>S: Draft notes/mcp-filesystem-nobara.mdx
  D->>S: Approve note, push main
  S->>A: git push triggers CI
  A-->>D: Build OK, SEO audit pass
  Q->>D: Weekly digest — 2 drafts ready, 3 FAQ ideas from GUIDE

Directive files (full agent specs)

AgentFile
CHORUSdirectives/CHORUS_portfolio_crew_orchestrator.md
HAWKdirectives/HAWK_competitor_intelligence.md
NEXUSdirectives/NEXUS_mcp_bridge_agent.md
FORGEdirectives/FORGE_code_execution_sandbox.md
AEONdirectives/AEON_ci_automation_agent.md
GUIDEdirectives/GUIDE_portfolio_concierge.md
QUILLdirectives/QUILL_chronicler_agent.md

Quick invoke reference

@ZED     — Route any request; assigns specialists
@HAWK    — Competitor intel for [region/vertical]
@CHORUS  — Generate case study for [project/industry]
@NEXUS   — MCP bridge for [IDE] ↔ [local capability]
@FORGE   — Verify code/commands before publish
@AEON    — Configure or run CI/SEO audit
@GUIDE   — Update concierge KB / embed on site
@QUILL   — Chronicle session; draft blog post

Blog answer (explicit)

Will you be writing blogs about what you work on?

Yes. QUILL exists for that. Every meaningful build session should produce at least one of:

  1. A session chronicle (internal, for you and the swarm)
  2. A draft note on davidcole.cloud (public, after your review)
  3. A case study (CHORUS → /projects) when the work is client-facing portfolio material

The site’s /notes section is the public face of your technical sandbox — agent workflows, homelab, and DIY infrastructure are all first-class content, not an afterthought.