Department Overview

All active departments, their recurring tasks, and current automation coverage. Click any card to view its full workflow.

Content

Blog pipeline, NLP-first workflow, client knowledge graphs, content strategy, AI-SEO optimization

49
blogs/month (24 clients)

Technical SEO

Health scans, crawl audits, Core Web Vitals, schema validation, sitemap monitoring

3x/wk
Tue/Thu/Sat 7 AM scans

Link Operations

LinkLab orders, Adsy vetting, backlink verification, link sheet management, live link checks

100
domains vetted daily @ 5 AM

Web Dev

Client site builds, scroll-stop animations, Cloudflare Pages, React/Vite/Tailwind/shadcn stack

10
active website projects

Optimization

OTTO recommendations, client improvement scans, conversion rate optimization, on-page audits

3x/wk
discovery & audit cycles

QA & Review

Self-review of correction rules, skill usage audit, process drift detection, quality gates

2x/wk
Wed/Sat 6 AM reviews

Blog Pipeline

NLP-first workflow. No blog gets written until NLP terms are discovered. Every gate must pass before the next step begins.

Hard Gate (blocks)
Skill loaded
Memory file read
Output produced
01
Topic Assignment
the operator provides: client name, article topic, 3 target keywords, geolocation
content-blogs.md blog-rules.md
  • No SMART sheets — topics come directly from the operator
  • The agent updates internal blog queue directly
  • Check if topic already exists (avoid cannibalization)
v
02
Load Client Knowledge Graph
Read client's brand voice, content rules, SEO context, visual identity before anything else
clients/{slug}-knowledge-graph.md client-knowledge-graph
  • Section 4: Brand Voice & Tone (mandatory)
  • Section 7: Content Rules (mandatory)
  • Section 2: Products & Services (for internal links)
  • Section 6: SEO Context (for keyword strategy)
v
!!
GATE: NLP Terms Discovery
Pull NLP terms for ALL blogs BEFORE writing content. Blog CANNOT proceed without NLP terms.
HARD GATE ai-seo
  • April 2026+: NLP-first is mandatory for every blog
  • Pull semantic terms, entities, related topics
  • These terms guide the entire content structure
  • NO writing begins until NLP terms are confirmed
v
03
SERP Analysis & Benchmarking
Check top 5 SERP results for target keyword — determine word count, content type, angle
content-strategy blog-rules.md
  • Average word count of top 5 results = benchmark
  • Identify content type (comparison, how-to, pillar, etc.)
  • Find differentiation angle (don't rewrite page 1)
  • Check for keyword cannibalization with service pages
v
04
Internal Link Planning
Analyze ALL product/pillar pages on client site. Map internal links before writing.
MUST COMPLETE BEFORE WRITE
  • Primary product page link (mandatory)
  • 1-2 related product pages
  • 1-2 related existing blog posts
  • No external hyperlinks — cite by name only
v
05
Load Writing Skills
Consult 4 mandatory skills before pen hits paper
ai-seo copywriting content-strategy schema-markup
  • ai-seo: LLM visibility, extractable passages, FAQ structure
  • copywriting: CTAs, persuasion elements, conversion copy
  • content-strategy: topic cluster alignment, pillar connections
  • schema-markup: BlogPosting + FAQPage structured data
v
06
Write Blog Content
Draft the blog using NLP terms, brand voice, SERP benchmark, and skill guidance
blog post draft Sonnet agent
  • Match word count to SERP benchmark + content type
  • First paragraph answers the query (40-60 words)
  • No filler, no stock AI phrases, no keyword stuffing
  • E-E-A-T signals: author byline, credentials, sources
  • YMYL topics: authority signals > keyword optimization
  • Named researchers + journal citations with years
  • FAQ section (3-5 questions) at bottom
  • CTA mandatory on every post
v
07
Image Generation
Generate hero + supporting images. Named {client}-{topic}-{type}.webp
1800x1013 WebP q92 tools-apis.md (image gen)
  • Hero/featured image near top, matches H1
  • Supporting images every 300-500 words
  • File name: {client}-{topic}-{type}.webp (3-7 words)
  • NOT blog-ID numbers in filenames
  • Generate with Nano Banana Pro
  • Compress under 100KB, descriptive alt text
v
08
Save to System & Supabase
Save blog to system (not chat). Update Supabase tracking. Don't dump in chat.
Supabase record saved file
  • Blog saved to filesystem, never dumped in Slack/chat
  • Supabase record created with metadata
  • Schema markup included (BlogPosting + FAQPage)
  • Meta description, featured image, alt text verified
v
!!
GATE: No Auto-Deploy
Present to the operator for approval. Never auto-publish or auto-deploy blog content.
APPROVAL REQUIRED

Feature Development Workflow

Superpowers-driven development. Strict gates at every phase — no skipping design, no code before tests, no completion without verification.

Hard Gate
Sub-agent dispatched
Decision point
01
Brainstorming (9 phases)
Explore context, ask clarifying questions, propose approaches, design in sections, write spec
superpowers/brainstorming docs/superpowers/specs/
  • Phase 1: Explore project context (files, docs, commits)
  • Phase 2: Offer visual companion if needed
  • Phase 3: Ask clarifying questions — ONE at a time
  • Phase 4: Propose 2-3 approaches with trade-offs
  • Phase 5-6: Present design in sections, get approval
  • Phase 7: Spec self-review (placeholders, consistency)
  • Phase 8: the operator reviews written spec
  • Phase 9: Hand off to writing-plans
v
!!
GATE: Design Approved
No implementation without written design + user approval. No exceptions.
HARD GATE
v
02
Writing Plans
Map file structure, break into 2-5 min tasks, include complete code in every step
superpowers/writing-plans docs/superpowers/plans/
  • Map file structure with clear boundaries
  • Bite-sized tasks (each 2-5 minutes)
  • NO placeholders, TBDs, or "similar to Task N"
  • Complete code in every step
  • Exact commands with expected output
  • Self-review against spec before handoff
v
??
Execution Mode Decision
Choose between sub-agent driven (recommended) or inline execution
Option A: Sub-agents Option B: Inline
v
03
Implementation (per task)
For each task: dispatch implementer -> spec review -> code quality review -> mark complete
implementer agent spec reviewer agent quality reviewer agent test-driven-development
  • RED: Write failing test first (mandatory)
  • GREEN: Simplest code to pass test
  • REFACTOR: Clean up after green only
  • Spec reviewer checks compliance first
  • Code quality reviewer checks second
  • Both must pass before moving to next task
v
!!
GATE: Verification
No completion claims without fresh verification evidence. Run the command, read the output, confirm.
HARD GATE verification-before-completion
v
04
Finishing Branch
4 options: merge locally, push & PR, keep as-is, or discard
finishing-a-development-branch NEVER push to main
  • Always push to staging branch
  • the operator reviews before merge to main
  • Descriptive changelog commit messages
  • noindex on non-production environments

Debugging Flow

Iron Law: NO FIXES WITHOUT ROOT CAUSE INVESTIGATION FIRST. Systematic debugging with mandatory phases.

!!
STOP: No Quick Fixes
Red flags: "Quick fix for now", "Just try changing X", "Multiple changes at once"
IRON LAW
v
01
Root Cause Investigation
Read errors carefully, reproduce consistently, check recent changes, trace data flow
systematic-debugging
v
02
Pattern Analysis
Find working examples, compare against references, identify differences, understand dependencies
v
03
Single Hypothesis
Form ONE hypothesis. Test minimally. Verify before continuing. Never multiple changes at once.
v
04
Implement Fix (TDD)
Create failing test case -> implement single fix -> verify fix -> done
test-driven-development 3+ failures = architectural problem

Daily Operations

Morning-to-evening task flow. Times in MST (America/Phoenix). Runs Mon-Fri unless noted.

5:00 AM — Daily
Link Potentials Vetting
Vet 100 domains (50 LinkLab + 50 Adsy). 1-hour timeout. Runs 7 days/week.
DAILY Link Ops
6:00 AM — Tue/Fri
Design Director Research
Browse Awwwards, Godly, 21st.dev, Dribbble for design inspiration and trends.
2x/WEEK Web Dev
6:00 AM — Wed/Sat
QA Self-Review
Review correction rules, audit skill usage, detect process drift. 15-min timeout.
2x/WEEK QA
7:00 AM — Mon/Wed/Fri
Client Improvement Scan
Scan 5 clients per run (15/week). Full cycle through all clients every ~2.3 weeks.
3x/WEEK Content
7:00 AM — Tue/Thu/Sat
Technical SEO Health Scan
Scan 5 clients per run. Crawl health, CWV, schema, sitemap checks. 20-min timeout.
3x/WEEK Technical
8:00 AM — Mon-Fri
Morning Check-In
Daily review: check system time, calendar for today, pending tasks, overnight results.
DAILY Operations
9:00 AM — Mon/Thu
LinkLab Order Check
Pull new links, update sheets, sync to Google Drive. 40-min timeout.
2x/WEEK Link Ops
9:00 AM — Tue/Thu/Sat
Client Optimization Discovery & Audit
Deep audit 3 clients per run (9/week). OTTO recommendations staged for approval.
3x/WEEK Optimization No auto-deploy
10:00 AM — Mon-Thu
Calendar Blocks Begin
Client 30-min blocks + website 2-hr blocks. Morning audits delivered to Slack before blocks start.
MON-THU Calendar
Weekly / Monthly Recurring
Sunday 3:00 AM
Backlink Verification
Verify all tracked backlinks are still live.
WEEKLYLink Ops
Sunday 4:00 AM
Blog Sitemap Scan
Check for new blog posts across all client sitemaps.
WEEKLYContent
1st & 15th @ 3:00 AM
Live Link Checker
Check ~242 tracked URLs for broken links. Biweekly.
BIWEEKLYLink Ops
1st @ 8:00 AM
Monthly Metrics & Wins
Compile monthly performance report across all clients.
MONTHLYOperations
15th @ 8:00 AM
Monthly Performance Follow-up
Mid-month check on metrics trends and client health.
MONTHLYOperations
25th
Blog Batch Pipeline
Monthly blog batch: prep + create 49 blogs. NLP terms, images, Supabase tracking.
MONTHLYContent
28th @ 10:00 AM
Monthly Retrospective
Review what worked, what didn't, process improvements for next month.
MONTHLYQA
Quarterly (Jan/Apr/Jul/Oct 1st)
Content Policy Scan
Audit content policies, compliance rules, industry regulation changes.
QUARTERLYQA

Automation Schedule

All cron-scheduled tasks by department. Each runs via crontab with gtimeout (macOS). Scripts in $REPO_ROOT/scripts/.

By Department

Content

Client Improvement Scan — MWF 7 AM (5 clients/run)
Blog Sitemap Scan — Sun 4 AM
Blog Batch Pipeline — 25th monthly

~18
runs/month

Technical SEO

Health Scan — TTS 7 AM (5 clients/run)

~13
runs/month

Link Ops

Link Vetting — Daily 5 AM (100 domains)
LinkLab Orders — Mon/Thu 9 AM
Backlink Verify — Sun 3 AM
Live Link Check — 1st & 15th 3 AM

~40
runs/month

Web Dev

Design Research — Tue/Fri 6 AM

~9
runs/month

Optimization

Discovery & Audit — TTS 9 AM (3 clients/run)

~13
runs/month

QA & Ops

QA Self-Review — Wed/Sat 6 AM
Morning Check-In — Mon-Fri 8 AM
Monthly Metrics — 1st 8 AM
Retrospective — 28th 10 AM
Policy Scan — Quarterly

~30
runs/month
Supporting Scripts

scan-blog-sitemaps.py

Crawls client blog sitemaps, detects new posts

pull-linklab-details.py

Pulls order details from LinkLab dashboard

update-backlink-sheets.py

Updates client backlink tracking sheets

check-live-links.py

Verifies ~242 tracked URLs are still live

Client Rotation & Calendar

31 active clients rotated Mon-Thu. Audits run morning-of, results delivered to Slack before the client's calendar block.

Priority (front of line)
Almost done
Waiting
Calendar Rules
C
Calendar Block Structure
Mon-Thu: 10 AM - 5 PM | Friday: 2 PM - 5 PM only | No weekends
  • Client blocks: 30 minutes each
  • Website builds: 2-hour blocks
  • Shuffle slots: 4x 30-min "Open Slot" per day
  • MFC Block: 1.5 hrs, Mon-Thu, 10 or 11 AM
  • Never conflict with the operator's existing events
  • Only manage agent-created events
Audit Alignment Flow
AM
Morning: Run Client Audit
Run audit for clients scheduled on today's calendar
v
DM
Deliver to Slack (DM Only)
Specific page URL + exact change + why it matters. Delivered BEFORE client block starts.
DMs only, NEVER threads
v
BK
Calendar Block: Execute
the operator uses pre-queued audit results during client's scheduled time
Priority Clients (Front of Line)

Client A

Sample priority client

Client B

Sample priority client

Website Build Queue (Priority Order)

1. Sample Client

Sample status

2. Sample Client

Sample status

3. Sample Client

Sample status

Billing-Based Review Calendar
8th of month
Sample Client
Billed 1st — review 7 days later
10th of month
Sample Client
Billed 3rd

Content Department

Blog production, content strategy, knowledge graphs, AI-SEO optimization. 49 blogs/month across 24 clients.

Recurring Automation
Mon/Wed/Fri 7:00 AM
Client Improvement Scan
5 clients per run. Content quality, gaps, recommendations. Full cycle ~2.3 weeks.
Sunday 4:00 AM
Blog Sitemap Scan
Check all client sitemaps for newly published posts. Script: scan-blog-sitemaps.py
25th of month
Monthly Blog Batch
Prep + create 49 blogs. NLP terms pulled, images generated, Supabase records created.
Skills Used

ai-seo

LLM visibility, extractable passages, FAQ structure

copywriting

CTAs, persuasion, conversion copy

content-strategy

Topic clusters, pillar connections

schema-markup

BlogPosting + FAQPage structured data

client-knowledge-graph

Brand voice, content rules, SEO context

Full Blog Pipeline

See Blog Pipeline view for the complete step-by-step flow with gates.

Technical SEO Department

Crawl health, Core Web Vitals, schema validation, sitemap monitoring. 5 clients per scan, full cycle ~2.3 weeks.

Recurring Automation
Tue/Thu/Sat 7:00 AM
Technical SEO Health Scan
5 clients per run, 20-min timeout. Crawl audits, CWV checks, schema validation, sitemap health.
Audit Checklist (Per Client)
1
Crawl Health
Check indexing status, crawl errors, robots.txt, canonical tags
v
2
Core Web Vitals
LCP, FID/INP, CLS measurements. Flag any regressions.
v
3
Schema Validation
Verify structured data is valid and present on key pages
v
4
Deliver Actionable Fix
Specific page URL + exact change + why it matters -> Slack DM

Web Dev Department

Client website builds. React/Vite/Tailwind/shadcn stack. Cloudflare Pages deployment. 10 active projects.

Recurring Automation
Tue/Fri 6:00 AM
Design Director Research
Browse Awwwards, Godly, 21st.dev, Dribbble for inspiration and trends.
Deployment Rules
!!
NEVER Push to Main
Always push to staging. the operator reviews. Merge to main for production.
v
!!
Cloudflare: Git-First Only
No Direct Upload. No duplicate projects. example-preview.com = noindex dev domain.
v
OK
Footer: "Site by your agency"
Required on all client sites. noindex on non-production environments.
Available Skills

scroll-stop-builder

Apple-style scroll animations

site-deconstructor

Reverse-engineer existing sites

html-to-elementor

Convert HTML to Elementor JSON

3d-asset-generator

3D models & renders for web

remotion

Programmatic video with React

shadcn-ui

Component library builder

kling-prompt

AI video prompt generator

frontend-design

Modern frontend interfaces

Optimization Department

OTTO recommendations, on-page audits, conversion optimization. 3 clients per run, full cycle ~4 weeks.

Recurring Automation
Tue/Thu/Sat 9:00 AM
Client Optimization Discovery & Audit
3 clients per run, 30-min timeout. Deep optimization analysis.
NEVER auto-deploy OTTO
Critical Rule
!!
NEVER Auto-Deploy OTTO Optimizations
Always present recommendations for the operator's approval first. No auto-push.
CRO Skills Available

page-cro

Landing page conversion optimization

form-cro

Form conversion optimization

popup-cro

Popup & modal optimization

onboarding-cro

User onboarding flow optimization

QA & Review Department

Self-correction, process drift detection, skill usage auditing. Keeps all other departments honest.

Recurring Automation
Wed/Sat 6:00 AM
QA Self-Review
Review correction-rules.md, audit skill usage across recent tasks, detect process drift.
28th @ 10:00 AM
Monthly Retrospective
What worked, what didn't, process improvements for next month.
Quarterly (Jan/Apr/Jul/Oct)
Content Policy Scan
Audit industry regulations, content compliance, policy changes.
QA Checks Applied to All Workflows
1
Correction Rules Check
Scan memory/correction-rules.md. Verify recent work didn't violate any rules.
v
2
Skill Usage Audit
Did recent tasks consult skills before executing? Flag any that didn't.
v
3
Process Drift Detection
Compare actual behavior against defined workflows. Flag deviations.
Non-Negotiable Gates (Applied Everywhere)

Design Gate

No implementation without written design + user approval

Test-First Gate

No production code without a failing test first

Verification Gate

No completion claims without fresh verification evidence

Root Cause Gate

No fixes without root cause investigation first

Review Gate

No merge without code review (spec + quality)

Approval Gate

No auto-deploy anything — the operator approves first