Execution Plan — April 2026

90-Day Transformation Execution Plan

7 gaps × 13 actions + 7 workstreams → sequenced into weekly sprints with time allocations. Target: agent fleet operational by July 2026.

190-Day Execution Plan

2Workstreams × Gap Correlation

3Managed Agents Deployment

4UAT Automation

Strategic Advice

You own 24 of 62 tasks but most of your work is decisions and reviews, not execution. Total Jake commitment: ~57 hours across 12 weeks (~4.5 hrs/week). Total delegated: ~117 hours across your team.

Gaps 2, 5, and 6 run in parallel — different owners (you, SEO Lead, PM). Assign Day 1 and they execute independently.
Your highest-leverage hours are system prompts and role cards. Nobody else can design agent architectures or redefine roles.
Use Claude to draft, you to decide. SOPs, role cards, one-pagers — Claude drafts in 20 min, you review in 15 min.
Don't touch G4 Hybrid OS until agents are running. Sprint 2 at earliest. Beta enrollment is a phone call, not a project.
Ship the positioning video last. You need case studies and agent results before you have a story worth telling.

Sprint 1 — Days 1–30

"Structure the Knowledge, Redefine the Roles"

Jake: ~24 hrsTeam: ~55 hrsGaps: G2 Knowledge Base (HIGH) · G3 Role Redefinition (HIGH) · G5 AEO Layer (MED) · G6 Velocity (MED)

Week 1Audit, Schema, Role Cards

G2 Knowledge Base

SOP Audit + Top 10 Selection + Conversion Schema

Jake: 3hwith Senior Team Lead

Pull full SOP inventory from Google Docs, ClickUp, Skills folder. Rank by execution frequency. Select top 10 for conversion. Design structured markdown schema (input/output fields, decision trees, QA gates). Brief Senior Team Lead on schema so they start converting independently.

Output: SOP inventory · Top 10 ranked · Markdown schema template · Senior TL briefed

G3 Role Redef.

Role Card Template + All 9 Role Cards

Jake: 5hSolo (Claude-assisted)

Design one-page role card template: agents managed, human judgment calls, QA responsibilities, escalation triggers. Draft all 9 role cards — Claude drafts each in ~15 min, you review ~15 min. Roles: SEO Lead, Content Lead, PM/Account Strategist, IT Lead, CRM/Ads Specialist, Web Designer, Developer, Senior Team Lead, Content Writer.

Output: Role card template · All 9 role cards drafted

G3 Role Redef.

Workshop Deck Preparation

Jake: 1.5h

Build Agent Manager shift workshop deck. Structure: why shifting → what changes per role → demo one agent workflow (SEO reporting) → Q&A. Schedule for Week 2.

Output: Workshop deck · Session scheduled

Delegated — Parallel Week 1

SEO Lead3hMap Semantic SEO Production workflow. Identify AEO gate insertion point. (G5 AEO Layer)

PM / Acct Strategist4hDefine velocity metrics. Create ClickUp custom fields for tracking. (G6 Velocity)

Week 2Workshop + Review Parallel Tracks

G3 Role Redef.

Team Workshop: Agent Manager Shift

Jake: 2hAll Team Members

Deliver workshop. Present shift, walk through role cards, demo one agent workflow. Record session for future onboarding.

Output: Workshop delivered + recorded

G3 Role Redef.

1:1 Role Validation Check-ins (5 of 9)

Jake: 2h5 team members × ~20 min

Validate each person articulates: which agents they manage, which decisions remain human, QA responsibilities. Adjust cards from feedback.

Output: 5 role validations done · Refinements noted

G5 AEO + G6 Velocity

AEO Gate + Velocity Schema Approval

Jake: 2hSEO Lead + PM

Review SEO Lead's AEO insertion point — approve or redirect. Co-author updated Pillar 3 SOP. Review PM's velocity definitions and ClickUp fields — approve schema. Green-light baseline measurement.

Output: AEO gate approved · Velocity schema approved · Both teams proceed independently

Delegated — Parallel Week 2

Senior Team Lead12hConvert SOPs 1–5 to structured markdown using Week 1 schema. (G2 Knowledge Base)

SEO Lead4hUpdate Pillar 3 SOP with AEO step. Add AI visibility score to monthly report template. (G5 AEO Layer)

PM / Acct Strategist6hMeasure velocity baseline for clients 1–8. Backfill historical data. (G6 Velocity)

Week 3SOP Validation + Remaining Check-ins

G2 Knowledge Base

Validate Converted SOPs 1–5

Jake: 2.5hwith Pillar Leads

Run each of the first 5 converted SOPs through Claude agent on test client scenario. Score accuracy (≥90% target). Flag revisions back to Senior Team Lead. ~30 min per SOP.

Output: 5 SOPs validated or sent back

G3 Role Redef.

Remaining 4 Role Validation Check-ins

Jake: 1.5h4 team members × 20 min

Complete remaining validations. Finalize all 9 role cards. Hand off to PM to update ClickUp role assignments.

Output: All 9 validations done · PM updates ClickUp · G3 Role Redefinition closed

Delegated — Parallel Week 3

Senior Team Lead12hConvert SOPs 6–10. Apply revision notes from 1–5 validation. (G2 Knowledge Base)

PM / Acct Strategist6hVelocity baseline clients 9–15. Build ClickUp velocity dashboard. (G6 Velocity)

SEO Lead4hTrain team on consensus audit Skill. Run AEO audit on all new content for validation week. (G5 AEO Layer)

PM / Acct Strategist2hUpdate ClickUp role assignments for Agent Manager definitions. (G3 Role Redefinition)

Week 4Sprint 1 Close + Architecture Sketch

G2 Knowledge Base

Validate SOPs 6–10 + Index All in ClickUp

Jake: 2.5hwith Pillar Leads

Same validation for SOPs 6–10. Once all pass, PM indexes in ClickUp SOP Library.

Output: All 10 SOPs validated + indexed · G2 Knowledge Base closed

G1 Agent Fleet

Agent Fleet Architecture Sketch (Sprint 2 Prep)

Jake: 2h

Map 6 pillars to agents: name, scope, Skills to wire, I/O schema outline, trigger conditions. Architecture sketch — detailed system prompts in Sprint 2.

Output: 6-agent architecture map · Sprint 1 scorecard reviewed

Delegated — Week 4

PM / Acct Strategist2hIndex all 10 knowledge bases in ClickUp SOP Library. (G2 Knowledge Base)

PM / Acct Strategist1hSet velocity reduction targets: 50% within 60 days. (G6 Velocity)

SEO Lead2hConfirm 100% of new content passed AEO audit for full week. G5 AEO Layer closed.

Sprint 2 — Days 31–60

"Deploy the Agent Fleet, Test Productization"

Jake: ~22 hrsTeam: ~40 hrsGaps: G1 Agent Fleet (HIGH) · G4 Hybrid OS (MED-HIGH) · G7 Outbound + Brand (MED)

Week 5Agent System Prompts — All 6

G1 Agent Fleet

System Prompts + I/O Schemas: All 6 Pillar Agents

Jake: 6h~1h per agent with Pillar Lead

Per agent (~60 min): draft system prompt with Claude (~20 min), refine with Pillar Lead (~15 min), wire to Skills (~15 min), define triggers + QA handoff.

1 Market Intel → Keyword Architect, AI Visibility · 2 GBP → Agency OS, AI Visibility · 3 Organic → Consensus Audit, Content Audit · 4 Ads → Copy Lab, Campaign Builder, Radius Optimizer · 5 Web → AI Carousel, AI Visibility · 6 CRM → GHL API, HL Assistant, Lead Engine

Output: All 6 system prompts · Skills wired · Triggers + QA handoff defined

Delegated — Week 5

IT Lead4hTechnical wiring: connect prompts to Skills infrastructure, test API connections. (G1 Agent Fleet)

PM / Acct Strategist2hSet up ClickUp Agent Registry structure. (G1 Agent Fleet)

Week 6Agent Testing + Outbound Flywheel

G1 Agent Fleet

Test 6 Agents on Client Scenarios (Round 1)

Jake: 3hwith Pillar Leads

1 test per agent (6 total, ~30 min each). Real client per pillar. Score output quality, accuracy, time savings. Flag prompt refinements. Pillar Leads run second round independently.

Output: 6 test runs scored · Refinement notes

G7 Outbound

Outbound → Content Flywheel + Case Study Agent

Jake: 3hContent Lead + IT Lead

Design n8n workflow: CRM closed deal → case study agent → content queue → social scheduler. Build case study agent — system prompt + input schema + output format. Test on 1 recent win. IT Lead wires n8n after.

Output: Workflow designed · Case study agent built + tested

Delegated — Week 6

Each Pillar Lead6h totalRun second test scenario per agent independently. Document results. (G1 Agent Fleet)

IT Lead6hWire n8n workflow end-to-end. (G7 Outbound + Brand)

Week 7Hybrid OS + Agent Registry

G4 Hybrid OS

Package Hybrid OS + Draft One-Pager

Jake: 3hIT Lead + Content Lead

Define package: GHL Snapshot components, agents included, dashboard scope, strategy call format, community structure. Draft one-pager with Claude — positioning, inclusions, $1,500–$2,000/mo pricing.

Output: Hybrid OS spec · One-pager draft

G1 Agent Fleet

Review All 12 Tests + Refine + Registry

Jake: 3hPillar Leads + PM

Review Pillar Leads' 6 independent test results. Fix 2–3 prompts needing refinement. Document all 6 agents in ClickUp Agent Registry.

Output: 12 tests reviewed · Prompts refined · Agent Registry v1 live

Delegated — Week 7

IT Lead5hBuild GHL Snapshot for Hybrid OS. (G4 Hybrid OS)

Content Lead2hPolish one-pager. (G4 Hybrid OS)

PM / Acct Strategist3hDocument Agent Registry. Build beta onboarding checklist. (G1 Agent Fleet + G4 Hybrid OS)

Week 8Beta Enrollment + Sprint Close

G4 Hybrid OS

Enroll 2 Beta Clients

Jake: 1.5hPM supports onboarding

Call 2 long-retention clients. Pitch Hybrid OS beta with one-pager. Get commitment. PM activates onboarding checklist.

Output: 2 beta clients enrolled

G7 Outbound

Review Case Studies + Sprint 2 Scorecard

Jake: 1.5h

Review Content Lead's 3 auto-generated case study drafts. Approve or revise. Confirm social pipeline connected. Sprint 2 scorecard.

Output: 3 case studies approved · Pipeline live · Sprint 2 closed

Delegated — Week 8

Content Lead6hGenerate 3 case study drafts. Connect social scheduler to LinkedIn + YouTube. (G7 Outbound + Brand)

PM / Acct Strategist2hActivate beta onboarding for both clients. (G4 Hybrid OS)

Sprint 3 — Days 61–90

"Measure, Iterate, Scale"

Jake: ~11 hrsTeam: ~22 hrsGaps: G1 Agent Fleet (HIGH) · G4 Hybrid OS (MED-HIGH) · G7 Outbound + Brand (MED)

Weeks 9–10Agent Review + Positioning

G1 Agent Fleet

Agent Fleet Health Report v1

Jake: 3hPillar Leads provide data

Compile: accuracy rate, time savings, error types, override frequency for all 6 agents. Calculate total delivery hours reduction vs. baseline. Write Health Report v1. Retrain prompts. Update Registry.

Output: Health Report v1 · Prompts retrained · Target: ≥85% accuracy, ≥30% hour reduction

G7 Outbound

AI-Native Positioning — Website + Sales

Jake: 3hContent Lead supports

Rewrite seonavigator.online About + services pages. Update sales deck and proposals with AI-native messaging. Brief Content Lead on 3 social posts.

Output: Website copy updated · Sales materials refreshed · Social briefs assigned

Delegated — Weeks 9–10

Each Pillar Lead6h totalCollect agent performance data for Health Report. (G1 Agent Fleet)

IT Lead3hApply prompt refinements and Skills wiring updates. (G1 Agent Fleet)

PM / Acct Strategist2hUpdate Agent Registry. Update sales proposals. (G1 Agent Fleet + G7 Outbound)

Weeks 11–12Beta Evaluation + Video + Close

G4 Hybrid OS

Hybrid OS Beta Evaluation + Go/No-Go

Jake: 2.5hPM provides data

Review beta feedback. Calculate delivery cost reduction vs. DFY. Decision gate: both retained + outcomes equivalent + cost ≥40% down = Go. Write memo. If Go: outline Tier 3 launch plan.

Output: Go/No-Go memo with data · If Go: Tier 3 launch plan

G7 Outbound

Positioning Video + Final Scorecard

Jake: 2.5h

Record 3-min positioning video: problem → solution → proof. One take, ship it. Final 90-day scorecard — score every metric, lessons learned, next 90-day plan.

Output: Video recorded · Scorecard complete · Next 90-day plan · All gaps closed or at Go/No-Go

Delegated — Weeks 11–12

PM / Acct Strategist4hCollect beta feedback. Prepare Go/No-Go data. (G4 Hybrid OS)

Content Lead4hPublish 3 social posts. Polish + publish positioning video. (G7 Outbound + Brand)

PM / Acct Strategist3hIf Go: build Tier 3 onboarding sequence from Jake's outline. (G4 Hybrid OS)

Time Budget Summary

~57h

Total Jake time

~117h

Total delegated team

~4.5h

Avg Jake hrs/week

12 wk

Full transformation

Sprint	Week	Gaps	Focus	Jake	Team
S1	W1	G2 Knowledge Base · G3 Role Redefinition	SOP Audit · Schema · 9 Role Cards · Workshop Prep	9.5h	7h
	W2	G3 Role Redefinition · G5 AEO Layer · G6 Velocity	Workshop · Check-ins · AEO + Velocity Approval	6h	22h
	W3	G2 Knowledge Base · G3 Role Redefinition	SOP Validation 1–5 · Remaining Check-ins	4h	24h
	W4	G2 Knowledge Base · G1 Agent Fleet	SOP Validation 6–10 · Agent Architecture Sketch	4.5h	5h
S2	W5	G1 Agent Fleet Architecture	System Prompts — All 6 Agents	6h	6h
	W6	G1 Agent Fleet · G7 Outbound + Brand	Agent Testing · Outbound Flywheel	6h	12h
	W7	G4 Hybrid OS · G1 Agent Fleet	Hybrid OS Package · Test Review · Registry	6h	10h
	W8	G4 Hybrid OS · G7 Outbound + Brand	Beta Enrollment · Case Studies · Sprint Close	3h	8h
S3	W9–10	G1 Agent Fleet · G7 Outbound + Brand	Agent Health Report · Positioning	6h	11h
S3	W11–12	G4 Hybrid OS · G7 Outbound + Brand	Beta Go/No-Go · Video · Final Scorecard	5h	11h

Team Load by Role

Role	S1	S2	S3	Total	Key Work
Senior Team Lead	24h	—	—	24h	SOP conversion (all 10 to structured markdown)
PM / Account Strategist	15h	7h	9h	31h	Velocity tracking, ClickUp config, Agent Registry, beta onboarding, feedback
IT Lead	—	15h	3h	18h	Agent wiring, n8n automation, GHL Snapshot, prompt refinements
Content Lead	—	8h	4h	12h	One-pager, case studies, social publishing, video production
SEO Lead	9h	—	—	9h	AEO gate, Pillar 3 SOP, team training, validation
Each Pillar Lead	—	6h	6h	12h	Agent testing, performance data collection
All Team Members	2h	—	—	2h	Workshop participation

After Week 12, the transformation infrastructure is built. ~57 hours of Jake-time and ~117 hours of team-time builds the full operating system. Everything after is iteration — refining Skills, expanding orchestration, scaling Hybrid OS if greenlit.

How Workstreams Map to the 90-Day Gaps

The 7 internal workstreams are the operational engine that runs alongside the 7 strategic gaps. Each workstream feeds into one or more gaps — its AI tasks become the execution layer for the transformation. Below, each workstream shows which gaps it directly supports, what week it activates, and which tasks are AI-automatable vs. human-owned.

Workflow Optimization + SOP System

W1–2 Priority▾

Maps to 90-Day Gaps

G2 Knowledge Base SOPs are the raw material for agent-ready knowledge bases. Sprint 1, Weeks 1–4.

G1 Agent Fleet Structured SOPs become the knowledge layer agents query. Sprint 2, Weeks 5–7.

G3 Role Redefinition RACI from SOPs defines human vs. agent ownership. Sprint 1, Weeks 1–3.

AI Tasks (Automatable)7

SOP document drafts from process descriptions8h/wk576

Master workflow map generation (v1)5h/wk392

RACI matrix generation from role descriptions3h/wk448

Top 10 SOP bottleneck identification2h/wk648

SOP gap analysis across existing docs3h/wk448

How-To video script generation4h/wk392

ClickUp template descriptions from SOPs4h/wk448

Human-Owned (Strategic)5

Workflow audit interviews with dept leads

Final SOP approval and sign-off

Video recording walkthroughs (screen + voice)

RACI ownership confirmation with each person

Template QA before publishing to Vault

Key Deliverables

Master workflow map (end-to-end)

SOP Library — docs + videos

ClickUp templates with SOPs attached

RACI baked into every template

Onboarding Simplification

W3–6 Standardize▾

Maps to 90-Day Gaps

G3 Role Redefinition Onboarding checklists encode the new Agent Manager role definitions. Sprint 1.

G2 Knowledge Base Welcome packs reference the structured SOP library. Depends on WS01 completion.

AI Tasks (Automatable)5

Onboarding checklist generation per role6h/wk729

ClickUp onboarding task auto-generation4h/wk512

Role-specific welcome doc + resource packs3h/wk448

Day 1 Slack welcome + channel routing auto2h/wk648

30/60/90 day check-in reminders2h/wk512

Human-Owned (Strategic)5

Day 1 in-person culture welcome (CEO or PM)

Buddy assignment — judgment call, personality fit

30/60/90 check-in conversations

Hiring manager W4 progress review

Culture immersion — values storytelling

Key Deliverables

Day 1 → Week 4 onboarding flow

Checklists: HR → PM → Department

ClickUp auto-task generation on hire

Role-specific welcome resource packs

Communication + Automation

W2–4 + W7▾

Maps to 90-Day Gaps

G6 Velocity Deadline automations + status updates accelerate time-to-deliverable. Sprint 1.

G7 Outbound + Brand Slack→ClickUp automation prevents work dying in chat. Sprint 2, Week 7.

AI Tasks (Automatable)6

Slack message → ClickUp task auto-creation10h/wk648

Deadline reminder automation sequences8h/wk729

Overdue notification auto-routing + escalation5h/wk576

Automated weekly status update generation6h/wk512

Slack channel structure audit + recommendations2h/wk448

Stripe + ClickUp + Slack integration (Zack)4h384

Human-Owned (Strategic)4

Slack channel rules agreement with team

Communication migration decision — CEO+PM

Escalation path definition for critical blockers

Communication culture coaching (1:1)

Key Deliverables

Slack channel structure + rules

Migration plan: ClickUp chat → Slack

Automations: tasks, reminders, status updates

Stripe + ClickUp + Slack integration (Zack)

Internal Brain System

W3–4 + W9▾

Maps to 90-Day Gaps

G1 Agent Fleet Internal Brain is the coordinator agent for all workstreams. Sprint 2–3.

G6 Velocity Proactive alerts and dashboards drive deployment speed. Sprint 1, Week 4.

AI Tasks (Automatable)6

Proactive deadline alert automation8h/wk729

Weekly project health digest generation6h/wk648

Automated meeting action item extraction4h/wk648

Blocker detection from ClickUp status changes3h/wk448

Real-time bottleneck dashboard auto-update3h/wk448

Meta Agent async status collection4h/wk294

Human-Owned (Strategic)4

Dashboard KPI selection and design

Escalation decisions when brain surfaces blockers

Strategic interpretation of weekly health reports

Meta Agent persona, scope, and name definition

Key Deliverables

Automated reminders: due dates, deliverables, blockers

Weekly project health digest (auto-compiled)

Real-time visibility dashboard

Meta Agent MVP (Phase 2 — W9+)

Recurring Tasks + Templates

W1–2 Highest▾

Maps to 90-Day Gaps

G2 Knowledge Base Templates are how SOPs become executable in ClickUp. Sprint 1, Weeks 1–4.

G1 Agent Fleet Template-based workflows become agent trigger conditions. Sprint 2.

G5 AEO Layer GBP + OnPage module templates include AEO audit step. Sprint 1.

AI Tasks (Automatable)6

ClickUp recurring task template generation10h/wk729

Template gap audit across departments4h/wk576

SOP-to-template conversion automation5h/wk512

GBP Intelligence module templates3h/wk392

OnPage Intelligence module templates3h/wk392

CSV import file generation for ClickUp bulk load2h/wk448

Human-Owned (Strategic)4

Dept lead review of each template before publishing

Former PM's incomplete templates — dedicated review session

Training team leads on Template Vault usage

Judgment on which templates need video walkthroughs

Key Deliverables

ClickUp templates for every recurring workflow

SOPs attached to each template

Template Vault organization inside ClickUp

GBP / OnPage / Citation module imports

Performance Framework

W3–7 Sequential▾

Maps to 90-Day Gaps

G3 Role Redefinition SOWs formalize the Agent Manager role boundaries. Sprint 1.

G6 Velocity KPI scorecards track deployment speed and agent performance. Sprint 1–2.

G4 Hybrid OS Performance data proves whether Hybrid OS delivers equivalent outcomes. Sprint 3.

AI Tasks (Automatable)6

Department KPI scorecard generation5h/wk576

Weekly KPI report auto-compilation6h/wk729

OKR draft generation from strategic goals4h/wk448

Department SOW drafts from role descriptions4h/wk448

Performance review template generation3h/wk512

OKR + KPI tracking cadence automation3h/wk448

Human-Owned (Strategic)5

OKR calibration sessions with department heads

Final performance review conversations (1:1)

Compensation decisions tied to KPI outcomes

Strategic OKR direction-setting (CEO)

SOW sign-off and boundary-setting with each dept lead

Key Deliverables

Department SOWs — responsibilities + boundaries

Department KPIs — monthly execution metrics

Quarterly OKRs — direction + growth

Performance Review System: skills + execution + mindset

Culture + Mindset Upgrade

W4–12 Sustained▾

Maps to 90-Day Gaps

G3 Role Redefinition "Ownership Operator" is the cultural shift that makes Agent Manager roles stick. Sprint 1.

G7 Outbound + Brand Culture storytelling feeds the positioning narrative. Sprint 3.

AI Tasks (Automatable)5

Ownership Operator training material drafts4h/wk392

Hire-to-culture criteria documentation2h/wk392

Culture guideline docs (expected behaviors)3h/wk392

Workshop agenda + slide deck generation2h/wk343

Behavioral expectation rubric for reviews2h/wk392

Human-Owned (Strategic)5

Live Ownership Operator workshops (CEO or PM-led)

CEO cultural storytelling — modeling publicly

Peer recognition moments (spontaneous, human)

Culture hiring interview panels

Behavior coaching 1:1 conversations

Key Deliverables

Ownership Operator training + workshops

Hire-to-culture criteria (structured)

Culture guidelines — expected behaviors

Behavior tied to performance reviews

Workstream × Gap Correlation Matrix

Workstream	G1 Agent Fleet	G2 Knowledge Base	G3 Role Redef.	G4 Hybrid OS	G5 AEO	G6 Velocity	G7 Outbound
WS01 Workflow + SOP	●	●●●	●●
WS02 Onboarding		●	●●
WS03 Communication						●●	●
WS04 Internal Brain	●●●					●●
WS05 Templates	●	●●●			●
WS06 Performance			●●	●		●
WS07 Culture			●●				●

●●● = primary driver · ●● = strong support · ● = contributes

The 7 workstreams are the internal operational backbone. The 7 gaps are the strategic transformation targets. They're not separate — every workstream directly feeds at least 2 gaps. Execute the workstreams and the gaps close automatically.

What Changes with Managed Agents

Claude Managed Agents replaces your current DIY agent infrastructure. Instead of building your own agent loops, sandbox, tool wiring, and error recovery, Anthropic runs all of that on their infrastructure. You define the agent (system prompt, tools, MCP servers, Skills), define the environment (packages, network), and start sessions. The agent runs autonomously for hours, persists through disconnections, and self-recovers from errors.

Your existing Claude Skills become the brain. Skills like Keyword Architect, Radius Optimizer, Consensus Audit — these wire directly into Managed Agent definitions. No rebuild needed.
MCP servers you already use connect natively. ClickUp, Slack, GHL, Google Drive, Figma — Managed Agents supports MCP out of the box. Your current 6-MCP architecture plugs right in.
Multi-agent coordination (research preview) is exactly your 6-pillar fleet model. A coordinator agent can delegate to specialist agents — this is the Agency OS pattern at infrastructure level.
Pricing: standard API tokens + $0.08/hr active runtime. No idle-time charges. A 4-hour SEO audit that runs overnight costs ~$0.32 in runtime on top of token costs.
This replaces Gap 1 (Agent Fleet Architecture) entirely. The infrastructure Anthropic now hosts is what you were going to build yourself.

Architecture: Before vs. After Managed Agents

Before — Current DIY Stack

Individual Claude Skills triggered manually
Custom n8n workflows for each automation
No persistent sessions — each call is stateless
No error recovery — failures require human restart
No multi-agent coordination
Context window management is manual
~2–3 months to wire each new agent

After — Managed Agents

Agents defined by YAML or natural language
Anthropic handles sandbox, auth, tool execution
Persistent sessions across hours/days
Auto error recovery + checkpoint/resume
Multi-agent coordination (coordinator → specialists)
Built-in prompt caching + compaction
~1–2 days per agent deployment

Agent

Model + system prompt + tools + MCP servers + Skills. Created once, referenced by ID.

Environment

Cloud container with packages (Python, Node.js), network rules, mounted files. Your sandbox.

Session

Running agent instance. Persistent file system, conversation history, hours-long execution.

Events

Messages between your app and agent. User turns, tool results, status updates. All persisted.

6 Department Agents — Managed Agent Definitions

SEO Sentinel

37h/wk saved▾

Managed Agent Configuration

ModelClaude Opus 4.6 SkillsAgency OS · Content Audit · AI Visibility MCP ServersSEO Utils · Google Drive · ClickUp · Slack EnvironmentPython 3.12 + Node.js 22 · Network: SEO Utils API, GSC API, QueryMind Agent ManagerSEO Lead (human gate: strategy sign-off, client delivery) Session ModelScheduled daily: rank monitoring + alerts. On-demand: audits, briefs, reports

Automated by Managed Agent8

Monthly SEO performance reports10h/wk729

Technical audit data collection & crawl7h/wk648

Position tracking & rank monitoring alerts5h/wk648

Topical map & keyword clustering4h/wk448

AI Overview consensus content audit3h/wk392

Competitor AI visibility analysis3h/wk448

Schema markup audit & generation2h/wk448

GSC anomaly detection & weekly alerts2h/wk648

Remains Human-Owned6

SEO strategy & direction decisions

Client presentations & data interpretation

Algorithm penalty recovery decisions

Competitive positioning calls

Link building strategy & partner negotiations

Final content calendar approval

Deployment approach: Define agent via API with system prompt from existing SEO Sentinel skill. Mount SEO Utils MCP + Google Drive MCP. Schedule daily session for rank monitoring (runs ~20 min overnight). On-demand sessions triggered by ClickUp task status change via n8n webhook. Deploy time: 1 day.

Content Catalyst

30h/wk saved▾

Managed Agent Configuration

ModelClaude Sonnet 4.6 (speed-optimized for volume) SkillsAgency OS · Content Audit · Lead Engine · Consensus Audit MCP ServersGoogle Drive · ClickUp · Slack EnvironmentPython 3.12 · Network: QueryMind API, web search Agent ManagerContent Lead (human gate: quality review, brand voice, publish) Session ModelOn-demand: triggered per content brief. Batch: weekly meta optimization runs

Automated by Managed Agent7

Content brief generation (Koray EAV method)15h/wk729

Meta title & description batch optimization4h/wk576

Research & citation compilation4h/wk448

AI Overview consensus scoring3h/wk392

Internal linking map generation3h/wk392

Social post repurposing from articles3h/wk448

EAV/semantic entity gap identification2h/wk336

Remains Human-Owned5

Content strategy & editorial direction

Brand voice calibration per client

Final quality review before publish

Subject matter expertise injection

Experience-based content creation

Deployment approach: Define agent with Content Audit + Consensus Audit skills. Mount Google Drive MCP for brief templates. Session triggered per ClickUp content task. Agent generates brief → runs AEO audit → outputs to Google Drive → pings Content Lead in Slack for review. Deploy time: 1 day.

Revenue Relay

28h/wk saved▾

Managed Agent Configuration

ModelClaude Sonnet 4.6 SkillsLead Engine · HL Assistant · GHL API · Agency OS MCP ServersGHL (GoHighLevel) · ClickUp · Slack · Gmail EnvironmentNode.js 22 · Network: GHL API, Twilio SMS API Agent ManagerCRM Lead (human gate: upsell conversations, deal negotiations, relationship) Session ModelContinuous: monitors pipeline. Triggered: new lead → follow-up sequence

Automated by Managed Agent7

Lead follow-up automation sequences (GHL)10h/wk648

Meeting notes & action item extraction4h/wk648

GHL workflow automation build & test5h/wk448

Review request automation sequences3h/wk512

Pipeline stage smart tag automation3h/wk448

SMS/email drip sequence copy drafts3h/wk392

Lead scoring & smart segmentation2h/wk336

Remains Human-Owned5

CRM strategy & pipeline architecture

Upsell conversations & deal negotiations

Client relationship & trust building

Strategic campaign decisions

Onboarding experience design

Deployment approach: Define agent with GHL API + HL Assistant + Lead Engine skills. Mount GHL MCP for pipeline read/write. Long-running session monitors pipeline for new leads → triggers follow-up sequence → drafts SMS/email → flags high-value leads to human. Deploy time: 2 days (GHL auth + sequence testing).

Ad Arbitrage

25h/wk saved▾

Managed Agent Configuration

ModelClaude Opus 4.6 (complex multi-step campaign work) SkillsAgency OS · AI Visibility · AI Carousel · Copy Lab · Campaign Builder · Radius Optimizer · Conversion Tracker · Remarketing Engine · Performance Auditor · Paid-Organic Crosswalk MCP ServersGoogle Ads API · ClickUp · Slack · Google Drive EnvironmentPython 3.12 + Node.js 22 · Network: Google Ads API, Meta Ads API Agent ManagerAds Lead (human gate: budget changes, campaign launches, creative direction) Session ModelScheduled weekly: performance reports. On-demand: campaign builds, audits

Automated by Managed Agent7

Performance report generation & analysis10h/wk729

Ad copy variation drafts (Google + Meta)5h/wk512

Competitor ad monitoring & SERP analysis4h/wk448

Audience research & targeting maps3h/wk343

A/B test result analysis & summaries3h/wk448

AI shopping carousel feed optimization3h/wk392

Budget pacing alerts & threshold reports2h/wk648

Remains Human-Owned5

Campaign strategy & budget allocation

Creative direction & visual asset briefs

Platform policy compliance review

High-spend optimization judgment

New channel strategy decisions

Deployment approach: This agent has the most Skills wired (10). Use Managed Agents' multi-agent coordination: define a coordinator agent that dispatches to specialist sub-agents (Copy Lab, Campaign Builder, Radius Optimizer, etc.) based on task type. Scheduled weekly session pulls Google Ads data → generates performance report → flags anomalies → drafts optimizations for human review. Deploy time: 2 days.

Build Bot

18h/wk saved▾

Managed Agent Configuration

ModelClaude Sonnet 4.6 SkillsAgency OS · AI Visibility · AI Carousel MCP ServersFigma · ClickUp · Slack EnvironmentNode.js 22 + Python 3.12 · Lighthouse CLI · Network: web fetch for audits Agent ManagerIT Lead (human gate: architecture decisions, security, UX strategy) Session ModelOn-demand: triggered per migration/audit task. Batch: schema audits

Deployment approach: Define agent with AI Visibility + AI Carousel skills. Environment includes Lighthouse CLI for Core Web Vitals audits. Sessions triggered per migration task or schema audit request. Outputs redirect maps, schema markup, QA checklists to ClickUp. Deploy time: 1 day.

PM Pulse

27h/wk saved▾

Managed Agent Configuration

ModelClaude Sonnet 4.6 SkillsAgency OS · HL Assistant MCP ServersClickUp · Slack · Google Drive · Gmail · Google Calendar EnvironmentPython 3.12 · Network: ClickUp API, Google Workspace APIs Agent ManagerPM / Account Strategist (human gate: scope negotiation, resource allocation, risk) Session ModelScheduled: daily status digests, weekly reports. On-demand: onboarding, SOPs

Deployment approach: This is the coordinator agent — the "Internal Brain" from WS04. Scheduled daily session pulls ClickUp task statuses across all departments → generates health digest → flags blockers → posts to #ops Slack channel. Monthly session compiles client reports. Also runs as the onboarding agent: new hire tagged → checklist auto-generated → Slack welcome fired. Deploy time: 1 day.

Deployment Timeline — Revised with Managed Agents

Impact on 90-Day Plan

Managed Agents compresses Sprint 2 (Agent Fleet deployment) from 4 weeks to ~1 week. The infrastructure you were going to build (sandbox, error recovery, state management, tool orchestration) is now handled by Anthropic. Your Sprint 2 work becomes: define agents → test → deploy. That frees up ~15 hours of Jake-time and ~20 hours of IT Lead time.

Day	Agent	Deploy Time	Owner	What Changes vs. Original Plan
31	PM Pulse (P6)	1 day	Jake + PM	Deploy first — it's the coordinator. Daily health digests start immediately.
32	SEO Sentinel (P1)	1 day	Jake + SEO Lead	Highest hours recovered. Overnight rank monitoring sessions start Day 33.
33	Content Catalyst (P2)	1 day	Jake + Content Lead	Content briefs + AEO audit run as single session. No separate AEO step needed.
34–35	Revenue Relay (P3)	2 days	Jake + CRM Lead	GHL MCP auth setup takes an extra day. Pipeline monitoring starts Day 36.
36–37	Ad Arbitrage (P4)	2 days	Jake + Ads Lead	Most complex (10 Skills). Multi-agent coordination: coordinator + sub-agents.
38	Build Bot (P5)	1 day	Jake + IT Lead	Simplest agent. Lighthouse CLI in environment. Schema audits on-demand.

8 days

All 6 agents deployed (was 4 weeks)

165h/wk

Total hours recoverable across fleet

$0.08/hr

Runtime cost (+ standard token pricing)

Multi-Agent Coordination — The Agency OS Pattern

Managed Agents' multi-agent coordination (research preview) maps directly to your Agency OS coordinator/specialist architecture. The PM Pulse agent becomes the meta-coordinator that can delegate to specialist agents:

COORDINATORPM Pulse receives task → identifies which specialist agent handles it → delegates → collects output → synthesizes SPECIALISTSEO Sentinel ← "Run technical audit for Modern Collision" → returns audit report SPECIALISTContent Catalyst ← "Generate 5 content briefs for Elite Finish" → returns briefs with AEO scores SPECIALISTAd Arbitrage ← "Weekly performance report for Procam Detailing" → returns report with optimizations SPECIALISTRevenue Relay ← "Activate follow-up sequence for 12 new leads" → runs GHL automation

Access note: Multi-agent coordination is currently in research preview. Request access to join. Your existing Agency OS skill already implements the coordinator/specialist pattern — Managed Agents gives it infrastructure-level support.

Claude Managed Agents is the infrastructure answer to Gap 1 (Agent Fleet Architecture). Instead of spending Sprint 2 building sandboxes and error recovery, you spend 8 days defining and deploying 6 agents on Anthropic's hosted infrastructure. The remaining Sprint 2 time shifts to testing, refinement, and Hybrid OS work — accelerating the entire 90-day timeline by 2–3 weeks.

UAT Automation Across All 7 Departments

We mapped every UAT checklist item across all 7 teams (334 total checks) against Claude Managed Agent capabilities. Result: 239 fully automated, 77 semi-automated, 18 manual-only. Each department gets a dedicated UAT agent triggered by ClickUp status change, gating publishing, and reporting to Slack. Total cost per full-site UAT: ~$9.65.

Department Teams

SEO

9 SOPs + QueryMind

66 · 20 · 1

CRM

34 · 7 · 1

Content

+QueryMind Planner/Jobs

38 · 9 · 3

Google Ads

25 · 6 · 1

Meta Ads

18 · 7 · 1

Web Design

30 · 22 · 10

28 · 6 · 1

■ Auto ■ Semi ■ Manual | SEO covers 9 SOP categories + QueryMind pipelines · Content includes QueryMind Planner + Jobs

SEO

SEO Team UAT — 87 checks

66 auto · 20 semi · 1 manual▾

Skills: Agency OS · Content Audit · AI Visibility · Consensus Audit · Koray City Page Auditor · AI Carousel Optimizer
Platforms: QueryMind (Topical Maps, Content Auditor) · SEO Utils MCP · GSC API · Local SEO Automation Toolkit
Structure: 9 SOP categories (monthly + special) + 2 QueryMind pipeline validations. Four agents triggered independently at different stages.

1. Content Creation CP — 17 checks → seo-uat-agentMonthly · Per page

Pre-Publish — On-Page + Schema (12)	Status	Agent Method
Topic validated against QueryMind coverage gaps + GSC opportunity	AUTO	Pulls GSC data + QueryMind gap_label, ranks by search volume × gap priority
URL structure (no duplicate/incorrect slugs)	AUTO	Crawls sitemap, checks duplicates, trailing slashes, casing
Title & Meta Description complete + unique	AUTO	Parses <title> + meta, checks length, flags duplicates
Heading hierarchy (H1→H2→H3)	AUTO	Validates single H1, sequential nesting, no skips
Sitemap.xml accessible + valid	AUTO	Fetches /sitemap.xml, validates XML, checks URLs return 200
Robots.txt not blocking important pages	AUTO	Parses Disallow rules vs. sitemap URLs
Canonical tags correct (self-referencing)	AUTO	Checks self-referencing canonical per page
Internal links working (no 404s)	AUTO	Crawls all links, reports 404s/500s with source
Images have alt text	AUTO	Parses <img> tags, flags missing/empty alt
Page speed acceptable (Lighthouse)	SEMI	Runs Lighthouse — human confirms threshold per project
Schema markup valid (JSON-LD)	SEMI	Validates JSON-LD syntax — human checks business logic
Index/noindex correct	AUTO	Checks meta robots + X-Robots-Tag headers

Post-Publish (5)	Status	Agent Method
Request indexing via GSC API + update ClickUp status	AUTO	GSC URL Inspection API → request indexing, update ClickUp task to "Index Requested"
Share published page on GBP	SEMI	Agent drafts GBP post copy + image — human publishes (no GBP write API)
Verify redirects & canonical in production	AUTO	Follows redirect map, checks 301 + canonical on live URL
CQS audit triggered on published content	AUTO	Chains to auditor-uat-agent → ContentAudit record created
Monitor ranking changes for published pages (7-day window)	AUTO	SEO Utils MCP, flags drops >3 positions vs. baseline

2. GBP Management CP — 4 checks → seo-ops-agentMonthly

GBP Monthly Operations (4)	Status	Agent Method
Check GBP performance (Maps, Local Pack, keyword movement)	AUTO	SEO Utils MCP grid reports, WoW rank delta, flags drops >3 positions → Slack summary
Optimize services, products & categories	SEMI	Agent audits GBP vs. competitors (web scrape), proposes changes — human applies in GBP dashboard
Manage GBP content (posts, photos, videos)	SEMI	Agent generates post copy from templates + Content AI — human publishes via GBP or GHL Social Planner
Monitor reviews status	AUTO	Checks GHL Reputation dashboard or scrapes Google reviews, flags new negatives → Slack alert. Chains to SNMS P6 Reviews AI.

3. GSC Management CP — 4 checks → seo-ops-agentMonthly

GSC Monthly Operations (4)	Status	Agent Method
Monitor performance & trends (clicks, impressions, CTR, position)	AUTO	GSC API → compares vs. prior period, flags anomalies >15% change → Slack + ClickUp
Check indexing & sitemap status	AUTO	GSC Sitemaps API + URL Inspection API for key pages, flags "Not Indexed" with reasons
Audit technical issues (404s, redirects, crawl budget)	SEMI	Agent crawls site (bash + curl), checks 404s, redirect chains, duplicate titles — human implements fixes
Monitor crawl errors (ongoing alerts)	AUTO	GSC crawl stats API, alerts on new 404s/5xx errors → Slack

4. GMC Management CP — 5 checks → seo-ops-agentMonthly · E-commerce clients only

GMC Monthly Operations (5)	Status	Agent Method
Sync Shopify inventory & Google Sheets	AUTO	Shopify API → export inventory → write to Google Sheets. Fully scriptable in managed container.
Check item availability (stock ≤0 detection)	AUTO	Queries Shopify inventory API, flags items with stock ≤0, cross-references GMC feed
Out of stock update (SKU reconciliation GMC ↔ Shopify)	AUTO	GMC Content API export → Shopify check → match SKUs → update GMC status → log to report sheet
Fix GMC errors (missing fields, policy violations)	SEMI	Pulls GMC diagnostics API, proposes fixes — human applies policy-related fixes
Product optimization (AI Carousel Optimizer)	AUTO	AI Carousel Optimizer skill: EAV descriptions, custom label mapping, title optimization. Batch 40 products/session.

5. Monthly Report — 3 checks → seo-ops-agentMonthly · Per project

Monthly Report Generation (3)	Status	Agent Method
Review task overview + embed ClickUp deliverables on CRM	AUTO	ClickUp MCP → query completed tasks, generate summary, embed in GHL custom field or Google Doc
Export rank tracking report (SEO Utils) & embed on CRM	AUTO	SEO Utils MCP → rank tracking data export → generate report (xlsx/HTML) → upload to Drive → embed link in CRM
Rerun grid reports, export & embed on CRM	AUTO	SEO Utils MCP → trigger grid re-run → wait for completion → export → embed. Chains with previous step for single-pass monthly report.

6. Citation Management CP — 2 checks → seo-ops-agentSpecial · Not monthly

Citation Operations (2)	Status	Agent Method
Audit current citations (NAP consistency across directories)	AUTO	Local SEO Automation toolkit: scrapes major directories, compares NAP, generates audit report
Submit new citations to directories	MANUAL	Requires CAPTCHA, email verification, human login on 20+ directories. Agent prepares data package only.

7. SEO NEO CP — 4 checks → seo-ops-agentSpecial · Per campaign

SEO NEO Operations (4)	Status	Agent Method
Project pre-assembly & setup (ClickUp + tool config)	SEMI	Agent creates ClickUp structure from template, populates config — human provides strategy inputs
Set up & run heatmap	SEMI	Agent triggers heatmap tool via API if available — human reviews visual output
Run SEO NEO campaigns (GBP Blast, Snipper, DAS, RD100)	SEMI	Agent orchestrates via tool APIs where available — human monitors execution
Monitor grid ranking changes post-campaign	AUTO	SEO Utils MCP → automated grid pulls + delta tracking vs. pre-campaign baseline

8. Website Audit — 6 checks → seo-ops-agentSpecial · New projects

New Project Audit (6)	Status	Agent Method
KW analysis (seed → clustering via QueryMind topical map)	AUTO	QueryMind pipeline: seed → expansion → clustering → CORE/OUTER. Chains to topical-map-uat-agent.
City analysis (local SEO grid per market)	AUTO	SEO Utils MCP + Local SEO Automation: city-level grid data, generates comparison report
Grid report + Whitespark/SEO Utils rank report	AUTO	SEO Utils MCP → automated grid pull + rank tracking export. Being consolidated into single report.
LLM visibility check (ChatGPT, Gemini, Perplexity)	AUTO	AI Visibility Audit skill: queries AI search engines for business presence, scores visibility
AI ranking check across AI search engines	AUTO	AI Visibility Audit skill: checks ranking position in AI-generated results
Analyze data & build SEO game plan	SEMI	Agent synthesizes all audit data into structured report with recommendations — human makes strategic decisions on priorities/budget

9. Google Optimization CP — 4 checks → seo-ops-agentSpecial · New projects only

GBP New Project Optimization (4)	Status	Agent Method
Foundation content setup + competitor analysis	AUTO	Local SEO Automation: scrapes competitor GBP data (categories, hours, services, reviews), generates gap report
GBP content preparation (posts, service descriptions, Q&A)	SEMI	Agent drafts all content from templates + Content AI — human reviews before publishing
GBP optimization (fields, categories, attributes)	SEMI	Agent generates optimized field values — human applies in GBP dashboard (no write API)
Final review & sign-off	SEMI	Agent generates checklist completion report — human strategic sign-off

QueryMind: Topical Map & Clustering — 18 checks → topical-map-uat-agentPipeline · On completion

CSI & Clustering Quality (10)	Status	Agent Method
CSI defined: CE, SC, CSI fields populated in workspace	AUTO	Queries workspace settings → checks CE/SC/CSI non-null
Seed keyword produces ≥50 expanded keywords	AUTO	Counts ClusterKeyword records → flags <50
No single-keyword clusters (min 3 per cluster)	AUTO	Queries cluster sizes → flags any <3
No mega-cluster (>30% of total in one cluster)	AUTO	Calculates max cluster % → flags >30%
All clusters have CORE/OUTER classification	AUTO	Checks cluster.type field → flags NULL
Cluster names are descriptive (not generic)	SEMI	Checks for generic patterns — human validates semantic quality
No orphan keywords (all assigned to a cluster)	AUTO	Counts keywords with NULL cluster_id = 0
Priority labels (P1–P4) assigned from gap analysis	AUTO	Checks priority field distribution, flags >50% NULL
CORE clusters cover central entity's primary topics	SEMI	Compares CORE names vs. CSI — human confirms relevance
CSV import handles edge cases (duplicates, encoding, empty rows)	AUTO	Uploads test CSV with known issues, checks import count

Coverage & Gap Tracking (8)	Status	Agent Method
Coverage baseline calculated (CORE published / total CORE)	AUTO	Validates coverage % against manual count
Content gaps identified (covered/gap/unique labels)	AUTO	Checks gap_label populated ≥80% of keywords
Published URLs linked to ClusterKeyword records	AUTO	Cross-references published_url with site crawl
Coverage % updates on new publish	AUTO	Simulates publish → checks recalculation fires
Visualization renders all clusters	AUTO	Checks Vue Flow node count matches DB cluster count
CORE clusters show correct keyword counts	AUTO	Compares UI badges against database counts
Gap keywords visually distinct from covered	SEMI	Checks CSS status classes — human validates clarity
Search volume + KD populated where available	AUTO	Checks search_volume NULL rate, flags >30% missing

QueryMind: Content Auditor & CQS Scoring — 20 checks → auditor-uat-agentPipeline · On completion

Pre-Audit Validation (4)	Status	Agent Method
Source URL accessible (200, content extractable ≥500 words)	AUTO	Fetches URL, checks status + text length
Content extraction successful (Readability output valid)	AUTO	Checks extraction output length and quality
SERP benchmark fetched (≥5 competitors if keyword provided)	AUTO	Counts SERP results + competitor scrape success
Degradation level recorded if services unavailable	AUTO	Checks ContentAuditStep.degradation_level

CQS Scoring (8)	Status	Agent Method
All 6 dimensions scored (CSI, CoR, Density, SRL, TF-IDF, EEAT)	AUTO	Counts ContentAuditScore records = 6
Each dimension score 0–10 range	AUTO	Validates score bounds on all 6 records
CQS formula correct: (CSI×0.25 + CoR×0.20 + Density×0.15 + SRL×0.10 + TF-IDF×0.10 + EEAT×0.20) × 10	AUTO	Recalculates from dimensions, compares against stored score
Weights sum to 1.0	AUTO	Sums weight fields = 1.00
CQS 0–100 + AI Citability 0–10 in valid range	AUTO	Validates bounds on both scores
Dimension weights match spec	AUTO	Checks each: CSI=0.25, CoR=0.20, etc.
Score stored on ClusterKeyword.cqs_score	AUTO	Checks ClusterKeyword updated
Re-audit shows delta from previous score	SEMI	Checks for prior audit, calculates delta — human validates

Report Quality (8)	Status	Agent Method
BEFORE/AFTER examples per dimension	AUTO	Parses report for before/after fields — flags missing
SRL transformations (CE as Agent rewrites)	AUTO	Checks SRL section, validates CE as grammatical agent
Headings marked [OK] / [CHANGE] / [NEW]	AUTO	Parses heading audit — all headings have marker
BLUF suggestions per H2 section	AUTO	Checks BLUF section ≥1 suggestion per H2
E-E-A-T blocks (4 dimensions with ready-to-paste content)	AUTO	Validates 4 EEAT sub-sections exist
Priority system (CRITICAL/HIGH/MEDIUM/BONUS/SKIP)	AUTO	Checks recommendations have priority labels
Report in target language	SEMI	Language detection — human confirms quality
TF-IDF term map with section assignments	AUTO	Checks term-to-section mapping exists with ≥10 terms

4 Agents — Independent Triggers

seo-uat-agentCP1 technical checks
Trigger: ClickUp → "Ready for UAT"
~5–10 min · $0.50/run

seo-ops-agentCP2–9 monthly + special
Trigger: Scheduled monthly or ClickUp → "Run Monthly SEO"
~15–25 min · $0.60/run

topical-map-uat-agentQueryMind TM pipeline
Trigger: Pipeline status = "completed"
~5–8 min · $0.45/run

auditor-uat-agentQueryMind CQS pipeline
Trigger: ContentAudit status = "completed"
~6–10 min · $0.65/run

CTN

Content Team UAT — 50 checks

38 auto · 9 semi · 3 manual▾

Skills: Agency OS · Content Audit · Consensus Audit · Lead Engine
Scope: Standard content publishing checks + QueryMind Content Planner brief validation + QueryMind Content Job pipeline (Quick Generate + From Topical Map dual-path). QueryMind orchestrates the 4-step brief pipeline and the 5-step content job pipeline with built-in degradation handling. Three agents triggered at different stages.

Section A — Content Publishing (12 checks → content-uat-agent)

Pre-Publish (9)	Status	Agent Method
No typos or grammar issues	AUTO	Claude reads full text, flags errors with corrections
Clear CTAs (Book / Contact / Sign up)	SEMI	Identifies CTA elements — human judges effectiveness
Content aligns with brand voice	SEMI	Compares against brand guide — human confirms
No placeholders (lorem ipsum, TBD)	AUTO	Regex + Claude scan for lorem ipsum, TBD, TODO
Internal & external links working	AUTO	Crawls all links, checks status codes
Layout correct (desktop & mobile)	SEMI	Screenshots at viewports — human reviews
Images / videos load properly	AUTO	Checks media src URLs for 200, dimensions >0
Legal pages available (Privacy, Terms)	AUTO	Checks footer links to /privacy-policy, /terms → 200
No duplicate content	AUTO	Similarity scoring against other site pages

Post-Publish (3)	Status	Agent Method
Validate content on production	AUTO	Re-runs pre-publish checks on production URL
Monitor user behavior (scroll, bounce)	MANUAL	Needs GA4/Hotjar data — human interprets
Optimize content for conversion	MANUAL	Strategic decision: performance data + creative judgment

Section B — QueryMind: Content Planner & Brief Pipeline (22 checks → planner-uat-agent)

Pipeline Step Validation (12)	Status	Agent Method
Topic research: CSI foundations populated (CE, SC, CSI, semantic frame, query fanout)	AUTO	Checks ContentBrief.csi_foundations JSON — all required fields non-null
Competitor analysis: ≥5 competitors scraped with EAV triples	AUTO	Counts competitor URLs in eav_matrix, EAV count ≥30
URR classification applied (UNIQUE/ROOT/RARE distribution)	AUTO	Checks URR labels, validates ROOT = highest count
H1 follows formula: CE + UNIQUE attribute + SC context	SEMI	Parses H1 against CSI — human validates creative quality
H2 headings map to ROOT attributes	AUTO	Cross-references H2s against ROOT attributes
H3 headings map to RARE attributes or FAQ	AUTO	Cross-references H3s against RARE + query fanout
BLUF per H2 section (≤50 words)	AUTO	Extracts BLUF, counts words, flags >50
RAG chunks: 200–500 words, no cross-references	AUTO	Counts words, scans for "as mentioned above"
All 9 brief sections populated	AUTO	Checks all 9 JSON columns — flags NULL
Content gaps prioritized P1–P4	AUTO	Validates content_gaps has priority labels
Copywriter checklist has 15 items	AUTO	Counts items in copywriter_checklist array
Degradation level recorded	AUTO	Checks ContentBriefStep.degradation_level

Brief Quality (6)	Status	Agent Method
Brief in correct target language	AUTO	Language detection vs. topical map language_name
UNIQUE differentiators ≥2 angles	AUTO	Counts items in unique_differentiators
TF-IDF terms + LSI keywords ≥10	AUTO	Checks keywords_and_terms non-empty
Internal links target other topical map keywords	AUTO	Cross-references linking suggestions vs. ClusterKeyword table
Quality metrics has target CQS per dimension	AUTO	Checks quality_metrics for 6 CQS dimensions
Brief is actionable for content writer	SEMI	Claude reviews for clarity — human confirms

Post-Approval Handoff (4)	Status	Agent Method
Approved brief creates job with source_type='topical_map'	AUTO	Checks ContentJob: source_type, content_brief_id fields
Job runs only steps 9–12 (not full 12-step)	SEMI	Checks step log — human confirms no unnecessary steps
Brief data injected into draft prompt	AUTO	Checks draft metadata for brief reference
Batch planning respects P1→P4 ordering	SEMI	Checks queue order vs. priorities — human reviews

Section C — QueryMind: Content Job Pipeline (16 checks → content-job-uat-agent)

Draft Quality (8)	Status	Agent Method
Topical Map path: only steps 9–12 executed	AUTO	Checks step log — steps 1–8 skipped for topical_map source
Draft incorporates all 9 brief sections	AUTO	Checks draft metadata for section references
Draft follows H1/H2/H3 from approved brief	AUTO	Parses headings, compares against article_structure
BLUF in each H2 (first 50 words = direct answer)	AUTO	Extracts first 50 words per H2, checks answer pattern
EAV triples from brief referenced in draft	SEMI	Scans for key entities — human confirms coverage
Humanization removes AI patterns	AUTO	Flags "in conclusion", "it's important to note", excess passive
Formatting follows brand guide	SEMI	Checks structure — human validates voice
QA catches placeholder content	AUTO	Regex for TBD, lorem, [insert]

Post-Publish Coverage (8)	Status	Agent Method
ClusterKeyword.published_url set + returns 200	AUTO	Checks field + fetches URL
ClusterKeyword.content_job_id linked	AUTO	Validates FK relationship
Coverage % recalculated	AUTO	Manual calc vs. TopicalMap.coverage_percentage
Published keyword shows green on map	AUTO	Checks status = 'published'
Internal links from brief in published content	AUTO	Fetches URL, checks for internal links to map pages
Published content passes SEO UAT	AUTO	Triggers seo-uat-agent on published URL
CQS audit triggered on published content	AUTO	Checks ContentAudit record created
Content quality acceptable (human final review)	MANUAL	Does content meet client quality standards

3 Agents — Independent Triggers

content-uat-agentTrigger: ClickUp → "Content Ready for Review"
~3–5 min · $0.30/run

planner-uat-agentTrigger: ContentBrief.status = "completed"
~8–12 min · $0.70/run

content-job-uat-agentTrigger: ContentJob.status = "completed"
~5–8 min · $0.50/run

CRM

CRM Team + SNMS 6-Pillar UAT — 42 checks

34 auto · 7 semi · 1 manual▾

Skills: GHL API · HL Assistant · Lead Engine · Pipeline Reactivation Engine · SNMS Lead Engine
Scope: Standard CRM UAT + full SNMS 6-Pillar GHL deployment validation (Lead Capture, NEPQ, Nurture, Upsell, VIP, Reviews)

Standard CRM — Lead Capture & Flow (11)	Status	Agent Method
All forms submittable (contact, booking, register)	AUTO	Submits test data to each form, checks 200 + redirect
Data mapping correct (name, phone, email)	AUTO	Submits payload → checks GHL contact via API → validates fields
No duplicate leads on re-submit	AUTO	Submits same lead twice → checks GHL for duplicates
Leads go to correct pipeline/stage	AUTO	Queries GHL pipeline API → confirms stage match
Workflows triggered on lead creation	AUTO	Checks GHL workflow execution log for test contact
Leads assigned to correct reps	AUTO	Queries contact assignment vs. routing rules
Email/SMS delivery successful	AUTO	Checks GHL conversation history for test contact
No delivery failures in logs	AUTO	Queries activity log for bounce/failure events
Retry mechanism works	SEMI	Simulates failure — retry window human-defined
Alerts for unassigned leads	AUTO	Creates unassigned contact → checks alert fires
DND / opt-out flags validated	SEMI	Checks DND flags — compliance needs legal review

P1 — Speed-to-Lead (<30 sec) (5)	Status	Agent Method
Primary Bot ("SNMS Lead Intake") in Auto-Pilot mode, 1s wait	AUTO	GHL API: checks bot config — mode=auto-pilot, wait_time=1
Chat Widget (All-in-One) installed on shop website	AUTO	Fetches shop URL, checks for GHL chat widget script in source
Bot responds to test message in <30 seconds	AUTO	Sends test SMS → measures response timestamp delta
WF-01 creates opportunity + auto-tags source	AUTO	Creates test contact → checks opp created + src_* tag applied
SLA timer fires internal notification at 5 min	AUTO	Creates contact with no bot engagement → checks notification log

P2 — NEPQ Qualification via Flow Builder (5)	Status	Agent Method
Flow Builder: 4 NEPQ nodes configured (Situation→Problem→Implication→Need Payoff)	AUTO	GHL API: checks bot flow structure — 4 Capture Information nodes exist
Custom fields populated (vehicle, pain_point, services)	AUTO	Runs test conversation → checks contact custom fields populated
AI Splitter branches correctly (Qualified → Book / Not → Nurture)	AUTO	Tests both branches: qualified contact gets booking link, unqualified enters WF-03
Appointment Booking action sends link to correct calendar	AUTO	Checks booking link URL matches service-specific calendar
Knowledge Base trained (website crawled + 10 FAQ + objection doc)	AUTO	GHL API: checks KB sources count ≥3, FAQ count ≥10

P3 — 90-Day Nurture (4)	Status	Agent Method
5 service-specific campaigns loaded (Stop on Reply = ON)	AUTO	GHL API: checks 5 campaigns exist with stop_on_reply=true
Re-Engagement Bot activates on nurture reply	AUTO	Simulates reply in nurture → checks Re-Engagement bot responds
Tag-based segmentation: hot/warm/cold applied by behavior	AUTO	Simulates reply + no-reply scenarios → checks correct tags applied
450+ O3 templates loaded in SMS + Email library	AUTO	GHL API: counts SMS + email templates, flags if <400

P4–P6 — Upsell, VIP, Reviews (10)	Status	Agent Method
Upsell WF-11 triggers on Won + service tag + correct wait days	AUTO	Simulates Won opp with svc_ppf → checks WF-11 fires at Day 30
Membership products configured in GHL Payments/Stripe	AUTO	GHL API: checks recurring products exist (Monthly Detail Club, Annual Plan)
B2B pipeline separate with 6 stages	AUTO	Checks pipeline config: 6 stages from Prospecting → Closed
VIP tier tags auto-applied (Bronze on 1st win)	AUTO	Simulates first Won → checks vip_bronze tag applied
3-year post-purchase WF-08 fires per service type	SEMI	Checks WF-08 structure has service-tag branching — human validates sequence timing
Win-back WF-09 triggers after 90-day inactivity	AUTO	Checks workflow date filter: last activity >90 days + has Won opp
Reviews AI Auto-Pilot active (4–5 stars, 2h delay)	AUTO	GHL API: checks Reviews AI config — mode=auto-pilot, star_filter=4-5, wait=2h
Negative reviews (1–3 stars) NOT auto-replied — task created for Sales Manager	AUTO	Simulates 2-star review event → checks task created, no auto-reply
Review request WF-07 fires after "Completed" stage	AUTO	Moves opp to Completed → checks SMS with review link sent after 2h
Referral trigger links working + attribution tags applied	AUTO	Clicks test trigger link → checks src_referral + referred_by_* tags

Post-Publish — Production Validation (7)	Status	Agent Method
Re-test all forms in production	AUTO	Same form tests against production URL
Full happy path: lead → AI → qualify → book → service → review → referral	SEMI	Agent runs automated flow — human monitors for edge cases
Full sad path: lead → no reply → nurture → cold → win-back	SEMI	Agent triggers flow — human validates timing/messaging quality
Voice AI Agent answers test calls (3+ scenarios)	SEMI	Agent dials test number — human evaluates conversation quality
Monitor first response time + contact rate (48h window)	SEMI	Agent pulls GHL data after 48h — human validates sample size
Stuck leads + automation failures checked daily	AUTO	Daily session: contacts stuck >24h in same stage
All 14 workflows firing correctly in production	MANUAL	Requires 5-day soft launch monitoring with real leads

Agent: crm-uat-agent · Opus 4.6 · ~15–25 min · ~$1.40/run

SkillsGHL API · HL Assistant · Lead Engine · SNMS Lead Engine · Pipeline ReactivationMCPGHL · ClickUp · SlackTriggerClickUp → "CRM Ready for UAT" → n8n → session (runs SNMS 6-pillar checks in addition to standard)

GAds

Google Ads + 8 Skills UAT — 32 checks

25 auto · 6 semi · 1 manual▾

Platform: Google Ads Orchestrator
Skills: Keyword Architect · Copy Lab · Campaign Builder · Radius Optimizer · Conversion Tracker · Remarketing Engine · Performance Auditor · Paid-Organic Crosswalk

Tracking + Tagging (8)	Status	Agent Method
Google Tag / GTM properly installed	AUTO	Checks page source for gtag.js or GTM container
Conversion: form submissions tracked	AUTO	Submits test form, checks dataLayer push / GA4 event
Conversion: button clicks tracked	AUTO	Simulates clicks, checks dataLayer events
Conversion: purchase/booking tracked	SEMI	Validates tracking code — actual purchase needs human
Remarketing tags active (Google Ads pixel)	AUTO	Checks for remarketing pixel in page source
Google Ads linked with Google Analytics	AUTO	Queries GA4 Admin API for linked accounts
UTM tracking correct on all ad URLs	AUTO	Parses ad URLs for utm params, validates consistency
GHL pipeline sync configured (GCLID tracking)	AUTO	Conversion Tracker skill: checks GHL webhook → Ads offline import setup

Keyword Architect + Copy Lab (8)	Status	Agent Method
Keyword clusters generated with match types assigned	AUTO	Keyword Architect output: validates clusters have broad/phrase/exact assignments
Negative keyword list created (shared + campaign-level)	AUTO	Checks negative list exists with ≥50 terms, no keyword cannibalization
Ad group structure follows keyword clustering (no cross-pollination)	AUTO	Validates each ad group's keywords belong to a single semantic cluster
RSA headlines: 15 per ad group, unique, within 30 chars	AUTO	Copy Lab output: counts headlines, checks char limits, flags duplicates
RSA descriptions: 4 per ad group, within 90 chars	AUTO	Counts descriptions, validates char limits
Ad strength rating ≥ "Good" on all RSAs	SEMI	Agent checks Ads API ad_strength field — human reviews if "Average"
Sitelinks, callouts, call extensions configured	AUTO	Copy Lab output: checks extension count ≥4 sitelinks + ≥4 callouts
A/B test framework defined (pin positions documented)	SEMI	Agent checks for pin strategy doc — human validates creative rationale

Campaign Builder + Radius Optimizer (8)	Status	Agent Method
Campaign structure matches Campaign Builder spec	AUTO	Compares live Ads structure against Campaign Builder output
Bid strategy correctly set (Maximize Conversions / Target CPA)	AUTO	Ads API: checks bid_strategy field matches spec
Budget allocation matches specified tier	AUTO	Sums campaign budgets, compares against total spend plan
Proximity/radius rings configured with bid modifiers	AUTO	Radius Optimizer output: validates tiered rings in Ads location targeting
Google Ads Scripts installed for radius automation	SEMI	Checks script exists in Ads account — human validates logic
Landing page loads fast (Lighthouse ≥70)	AUTO	Runs Lighthouse on all landing page URLs from campaigns
Ad scheduling matches business hours	AUTO	Ads API: checks ad_schedule settings against client hours
Remarketing audiences created (RLSA + Customer Match)	AUTO	Remarketing Engine: checks audience lists exist in Ads account

Post-Launch — Performance + Crosswalk (8)	Status	Agent Method
Conversions verified in real-time via GA4	SEMI	Queries GA4 Real-Time API — human confirms volume
Quality Score audit: all keywords ≥5	AUTO	Performance Auditor: pulls QS per keyword, flags <5
Wasted spend identified (search terms report audit)	AUTO	Performance Auditor: analyzes search terms, flags irrelevant with spend
Tracking not dropping (conversion count vs 7-day avg)	AUTO	Daily check: flags >30% conversion drop
Paid-Organic Crosswalk analysis generated	AUTO	Crosswalk skill: merges Ads data + organic rankings, identifies overlap
Cost savings opportunities identified from organic overlap	SEMI	Agent flags keywords ranking #1-3 organically with paid spend — human decides
Impression share >70% on brand terms	AUTO	Ads API: checks brand campaign impression share
Value-based conversion assignment per service type	MANUAL	Conversion Tracker: validates values match client revenue — requires business input

Agent: gads-uat-agent · Opus 4.6 · ~12–18 min · ~$1.10/run

SkillsKeyword Architect · Copy Lab · Campaign Builder · Radius Optimizer · Conversion Tracker · Remarketing Engine · Performance Auditor · Paid-Organic CrosswalkMCPGoogle Ads API · ClickUp · SlackTriggerClickUp → "Ads Ready for UAT" → n8n → session (runs full 8-skill validation)

Pixel + Events (6)	Status	Agent Method
Meta Pixel base code + PageView event active	AUTO	Fetches page, checks for fbq('track','PageView')
Lead event firing on form submit	AUTO	Submits test form → checks Lead event registered
Purchase/CompleteRegistration event	SEMI	Validates code exists — actual purchase needs human test
Domain verified in Business Manager	AUTO	Checks DNS TXT record or meta tag for FB verification
Aggregated Event Measurement configured	AUTO	Queries Business Manager API for AEM config
Custom conversions set with correct URL rules	AUTO	Pulls custom conversion list, validates URL rules

Campaign Accelerator — Pre-Launch (10)	Status	Agent Method
Buyer persona generated (VoC mining + competitor intel)	AUTO	Accelerator pipeline: checks persona doc exists with demographics + psychographics
4-Campaign architecture built (C-1 ASC+, C-2 Manual ABO, C-3 Creative Lab, C-4 Retargeting)	AUTO	Checks Meta Ads account for 4 campaigns matching architecture spec
6 Hook Archetypes applied across ad creatives	SEMI	Agent checks creative copy against hook patterns — human validates creative quality
Anchor Offer defined (irresistible lead magnet for detailing)	SEMI	Agent checks offer exists in ad copy — human validates offer strength
Messenger flow configured (if using Messenger ads)	AUTO	Lead Gen Engine: checks Messenger automation exists in business page
Instant Forms built with correct field mapping to GHL	AUTO	Lead Gen Engine: checks form fields + webhook/CRM integration active
Post ID strategy documented (graduation criteria)	SEMI	Checks for Post ID graduation doc — human validates criteria
Audience targeting matches buyer persona (interests, lookalikes, custom)	AUTO	Compares campaign audience settings against persona demographics
Pre-launch checklist passed (Diagnostics skill)	AUTO	Diagnostics: runs 5-layer check (Objective→Targeting→Creative→Bidding→LP)
Lead delivery to GHL pipeline working	AUTO	Submits test lead via Instant Form → checks GHL contact + opp created

Post-Launch — Performance + Optimization (10)	Status	Agent Method
Real-time event tracking validated	SEMI	Checks Events Manager test events — human confirms behavior
Lead events match CRM data (Meta → GHL count alignment)	AUTO	Cross-references Meta lead count with GHL contact creation timestamps
CPL within target range	SEMI	Performance Hub: pulls spend + leads — human judges vs. target CPL
Campaign scoring Green/Yellow/Red applied	AUTO	Performance Hub: scores each campaign against benchmarks
Budget allocation quadrant analysis (Stars/Hidden Gems/Cash Cows/Money Pits)	AUTO	Optimizer: maps campaigns to quadrant, flags Money Pits for review
Audience saturation check (frequency <3)	AUTO	Optimizer: checks frequency metrics, flags saturation
Creative fatigue detection (CTR declining >20%)	AUTO	Diagnostics: compares WoW CTR, flags declining creatives
Funnel analysis: lead → appointment → show → close	SEMI	Optimizer: pulls funnel data — human validates stage conversion accuracy
Phone call tracking matched (if call ads)	SEMI	Lead Gen Engine: checks call tracking integration — human confirms accuracy
90-day roadmap generated from performance data	MANUAL	Strategic planning requiring human judgment on scaling direction

Visual & Brand Consistency (10)	Status	Agent Method
UI matches Figma design (if Design-First)	SEMI	Figma MCP: export baseline → Playwright screenshot → side-by-side comparison
Colors consistent with brand palette	AUTO	Extracts CSS color values, compares against brand palette doc
Typography correct (font, size, line height)	AUTO	Audits computed styles vs. design system specs
Spacing & alignment consistent	SEMI	Checks CSS consistency — visual alignment needs human
Buttons/forms/cards consistent across pages	AUTO	Audits all component CSS classes for consistency
No visual mismatch between pages	SEMI	Multi-page screenshot comparison
Icon style consistent (no mixed libraries)	SEMI	Detects icon libraries in use, flags mixed sets
No placeholder assets (dummy images, stock IDs)	AUTO	Checks filenames/alt for "placeholder", "dummy", "sample"
Images high quality (not blurry/upscaled)	SEMI	Checks resolution vs. display size — human reviews quality
Design token consistency (CSS variables match Figma variables)	AUTO	Figma MCP: get_variable_defs → compare against CSS custom properties

Playwright — Visual Regression (6)	Status	Agent Method
Full-page screenshot regression test passes (≤1% diff)	AUTO	npx playwright test visual — fullPage comparison against baseline
Per-section regression: Hero, Services, Testimonials, CTA, Contact, Footer	AUTO	Per-section element screenshots vs. baselines
Desktop (1440px) visual test passes	AUTO	Playwright project: Desktop Chrome viewport
Mobile (375px) visual test passes	AUTO	Playwright project: iPhone 14 viewport
Tablet (768px) visual test passes	AUTO	Playwright project: iPad viewport
Figma baseline comparison (Design-First only)	SEMI	Claude multimodal: reads Figma export + implementation screenshot, reports diffs

Playwright — Responsive Layout (8)	Status	Agent Method
Mobile: nav shows hamburger, cards stack vertically	AUTO	Playwright: checks #menu-toggle visible, nav hidden, cards Y-stacked
Desktop: nav shows links, cards in grid	AUTO	Playwright: checks nav visible, cards same-Y position
No broken layout on smaller screens (no overflow)	AUTO	Detects horizontal scroll, text overflow, clipping
Text readable on all devices (≥14px mobile)	AUTO	Checks computed font-size, WCAG AA contrast
CTA buttons mobile-friendly (tap target ≥48×48px)	AUTO	Validates tap target sizes per Google guidelines
No dead-end pages (all pages have nav + CTA)	AUTO	Crawls pages, flags any without navigation links
Real device testing (beyond emulation)	SEMI	Emulated tests auto — real iOS/Android needs human with device
Cross-browser: Chrome, Safari, Edge	SEMI	Chromium auto — Safari/Edge need BrowserStack or manual

Playwright — Interactions + Accessibility (10)	Status	Agent Method
Mobile menu opens/closes correctly	AUTO	Playwright: click toggle → menu visible → click link → menu closes
Hamburger has aria-expanded attribute toggling	AUTO	Checks aria-expanded false→true→false on toggle
CTA buttons have hover styles (visual change)	AUTO	Screenshots default + hover state, validates difference
Contact form accepts input and validates	AUTO	Playwright: fills all fields, verifies values, screenshots filled form
Heading hierarchy correct (single H1, sequential H2→H3)	AUTO	Counts H1 (must be 1), validates heading sequence
All images have alt text or role="presentation"	AUTO	Iterates all <img>, checks alt or role attribute
Form inputs have associated labels	AUTO	Checks label[for=id] exists for each input
Keyboard navigation reaches all interactive elements	AUTO	Tab through all elements, counts focused interactive elements
Color contrast WCAG AA compliant	AUTO	WCAG contrast check on all text/background combinations
Font size accessible (≥12px minimum)	AUTO	Min font-size check across breakpoints

Build + Deploy Pipeline (8)	Status	Agent Method
Astro build compiles without errors (npm run build)	AUTO	Runs build command, checks exit code 0
GitHub Actions CI pipeline passes	AUTO	Checks workflow run status via GitHub API
Vercel/Netlify deployment successful	AUTO	Checks deployment status API, validates live URL returns 200
SEO: meta tags, sitemap.xml, robots.txt present	AUTO	Chains to seo-uat-agent for standard SEO checks
User journey smooth (landing → form → submit → thank you)	MANUAL	Requires user testing or session recording
No confusing steps in conversion flow	MANUAL	Agent maps flow but can't judge confusion
Collect feedback from client/stakeholders	MANUAL	Human communication — can't automate
Performance: Lighthouse ≥70 mobile, ≥80 desktop	AUTO	Runs Lighthouse via Playwright, validates scores

Post-Deploy — Production Monitoring (20)	Status	Agent Method
Live UI matches approved design	SEMI	Side-by-side screenshots — human sign-off
UI inconsistencies in production detected	AUTO	Re-runs all Playwright tests on production URL
Fix spacing/alignment issues	MANUAL	Requires dev work — agent flags only
Users understand flow (session recording)	MANUAL	Needs Hotjar/FullStory analysis
No unexpected drop-off	SEMI	Pulls GA4 funnel data — human interprets
CTA visibility strong	SEMI	Above-fold CTA check — human reviews impact
Touch interactions work on real devices	MANUAL	Touch/swipe needs real device testing
Identify UX improvement opportunities	SEMI	Analyzes data patterns — human decides priorities
Propose UI/UX enhancements	SEMI	Drafts recommendations — human reviews
No users completing key actions with friction	MANUAL	Requires real user testing

Pre-Publish — Infrastructure (4)	Status	Agent Method
Production env configured (domain, SSL)	AUTO	DNS resolution, SSL cert validity + expiry, headers
HTTPS working (no mixed content)	AUTO	Crawls pages, flags http:// resource loads
Env variables correct (API keys, DB)	SEMI	Tests API connectivity — human verifies secrets
No staging/dev configs remain	AUTO	Checks for staging URLs, debug flags, console.log

Pre-Publish — APIs & Auth (5)	Status	Agent Method
All APIs functioning	AUTO	Health-check each endpoint, check 200 + valid response
Timeout & error handling	AUTO	Sends slow/invalid requests, checks graceful handling
Retry logic works	AUTO	Simulates failure, monitors retry in logs
Auth (OAuth/token) works	AUTO	Tests valid + invalid creds, checks token refresh
Webhooks sending & receiving	AUTO	Triggers event, checks receipt on endpoint

Pre-Publish — User Flows + Security (8)	Status	Agent Method
Full user flow works (landing→form→submit→thanks)	AUTO	Puppeteer navigates full flow, validates each step
Form validation correct	AUTO	Empty/invalid/valid submissions, checks messages
Error messages clear	SEMI	Captures messages — human judges end-user clarity
reCAPTCHA working	SEMI	Checks script loads + renders — can't solve (by design)
No exposed API keys/secrets	AUTO	Scans source + JS bundles for key patterns
Input validation (XSS/SQLi)	AUTO	Submits common payloads, checks sanitization
Security headers (CORS, CSP)	AUTO	Checks CSP, X-Frame, X-Content-Type, HSTS
SSL/TLS secure	AUTO	TLS version, cipher suite, cert chain

Pre-Publish — Performance + Testing (7)	Status	Agent Method
Page load <3 seconds	AUTO	Lighthouse TTFB + FCP + LCP measurement
No JS/CSS errors	AUTO	Captures console errors via Puppeteer
API response time acceptable	AUTO	Times each call, flags >500ms
Images optimized	AUTO	Checks sizes >200KB, WebP/AVIF usage
Mobile / tablet / desktop testing	AUTO	Emulates 6+ viewports, captures errors
Cross-browser (Chrome, Safari, Edge)	SEMI	Tests Chromium — Safari/Edge need manual or BrowserStack
Error logging + alerts configured	AUTO	Triggers errors, checks logs appear in monitoring

Post-Publish (11)	Status	Agent Method
Server uptime	AUTO	Pings every 5 min, alerts Slack on downtime
Error logs with real traffic	AUTO	Groups by type/frequency, flags new errors
API failure rates	AUTO	Queries monitoring for failure %, alerts >1%
User journeys not broken	SEMI	Re-runs flows — edge cases need real monitoring
No unexpected crashes	AUTO	Monitors 5xx + process restarts
CRM receives data	AUTO	Test form → check GHL contact creation
Payment works in production	MANUAL	Real payment test needs human with test card
Webhooks not failing	AUTO	Monitors delivery logs for failures
Traffic spike handling	AUTO	Basic load test, checks response degradation
Memory / CPU stable	AUTO	Queries hosting metrics API
Error rate within threshold	AUTO	24h error rate vs. threshold (<0.5%)

Full-Site UAT: One Trigger, All 12 Agents in Parallel

When a client site reaches final staging, one ClickUp status change to "Full UAT" triggers all 12 agents via n8n. Each runs its checklist independently. The site only goes live when all departments pass.

Agent	Checks	Auto	Semi	Manual	Time	Cost
seo-uat-agent	17	14	3	0	~5–10 min	$0.50
seo-ops-agent	32	20	11	1	~15–25 min	$0.60
crm-uat-agent	42	34	7	1	~15–25 min	$1.40
content-uat-agent	12	7	3	2	~3–5 min	$0.30
gads-uat-agent	32	25	6	1	~12–18 min	$1.10
meta-uat-agent	26	18	7	1	~10–15 min	$0.95
web-uat-agent	62	30	22	10	~15–25 min	$1.30
it-uat-agent	35	28	6	1	~15–30 min	$1.20
topical-map-uat-agent SEO	18	15	3	0	~5–8 min	$0.45
planner-uat-agent CTN	22	18	4	0	~8–12 min	$0.70
auditor-uat-agent SEO	20	17	3	0	~6–10 min	$0.65
content-job-uat-agent CTN	16	13	2	1	~5–8 min	$0.50
TOTAL (12 Agents)	334	239	77	18	~30–40 min	~$9.65

Before — Manual UAT All Teams

Each dept runs their checklist manually
~2–4 hours per team per site
Total: ~15–25 hours per full site UAT (incl. content pipeline)
Inconsistent, no cross-team report
Post-publish monitoring ad-hoc
~15–25 hrs/week agency-wide

After — 7 Agents in Parallel

All 7 agents run simultaneously
~30–40 min total (parallel)
239 of 334 checks fully automated
Unified report in ClickUp + Slack
Post-publish daily × 14 days
~$9.65/run · human only for 18 manual items

Full-site UAT becomes a $9.65, 40-minute automated process instead of a 20-hour manual effort. 239 checks never get skipped. 77 semi-automated checks flag for human judgment. 18 manual items (UX testing, payment, user feedback) stay human because they should be.