The Content Factory is run by named agents and teams. This page records who owns what, which skills they use, what each executor may touch, how work is handed off, and which jobs wake on a schedule.
Agents, skills, tasks, and teams are different objects. An agent is a named executor or orchestrator. A skill is a versioned procedure it can use. A task is a bounded unit of work with acceptance criteria. A team coordinates agents. Lifecycle placement, permission tier, connector, schedule, and publication status are separate fields. See A Skill Is a Function. An Agent Is a Person.
40 roster records: 32 agents and 8 vertical skill variants; 31 records have published definitive articles (●). A published skill is ready to attempt in a configured environment; it is not thereby proven reliable or elevated above E0. Until a coming-soon article ships, use the bounded SOPs in the Task Library.
A qualifying public, privacy-safe run should become a short Meta Article linked to its definitive article so lessons flow back into the skill. That article is historical context, not formal E1–E4 evidence.
How the factory fits together
The Content Factory has four canonical production phases: Produce, Process, Post, and Promote. Plumbing plus Access & Governance are prerequisites. Reason / Plan guides decisions across all four phases. Perform closes the measurement loop. A required input such as the Topic Wheel may gate a specific downstream task without becoming a fifth phase.
A documented task becomes a named production role through the five-step agent-building loop.
Permissions — what each agent is allowed to touch
A skill is intent. A connector is capability. A human grants the connector. That separation is the whole permission model, and it is why a skill file is safe to publish in public: the file describes what to do, but it cannot do anything until someone attaches it to an account with the right connectors turned on.
Three tiers, and every agent on this page declares which one it needs:
| Tier | What it can do | Approval gate |
|---|---|---|
| Read | Web search, fetch public pages, read the WordPress REST search endpoint, read Drive/Gmail/Calendar, read Basecamp, read analytics. | None beyond connecting the account once. |
| Draft | Create drafts and internal records: WordPress drafts, Gmail drafts, Basecamp to-dos and threads, Drive files, Descript projects, Basecamp reassignment. | None per item, but the agent reports what it changed. |
| Publish | Anything the public or a client sees: publishing a post, sending an email, launching or scaling an ad, inviting a client to a project. | Explicit human go-ahead, every time. Agents draft and stop. |
Connector map
| Connector | Agents that need it | Highest tier used |
|---|---|---|
| WordPress (self-hosted, browser session + REST nonce) | The Publisher, Personal Brand Site Engineer, Adrian, Video Repurposing, BlitzAdmin fleet skills | Publish |
Meta Ads (mcp.facebook.com/ads) |
The whole Dollar a Day Boost Team, Facebook Ad Agent, all verticals | Publish |
| Basecamp (browser session) | The RACI Enforcer, Stephanie | Draft |
| Gmail / Google Calendar / Google Drive | Stephanie, Capture Coordinator, Library Curator, Memory Librarian | Draft |
| Descript | The Video Editor | Draft |
| Ahrefs / DataForSEO / Search Console | Christopher, Olivia, Redirect-Hygiene Auditor, SEO watch | Read |
| Claude in Chrome | Any agent reaching a site with no API — Basecamp, self-hosted WordPress, Knowledge Panel work | varies |
Scheduled jobs — what wakes up without being asked
Some agents run on a clock. These are live scheduled tasks on the account, not aspirations. Times are UTC.
| Job | Cadence | Agent it wakes | What it does |
|---|---|---|---|
| Weekly authority-citation sync — podcast cluster, counts, skill downloads | Mondays 11:37 | Ethan — Positive Mentions | Re-counts podcast appearances and citations, refreshes the proof library, tracks skill-pack downloads. |
| Weekly SEO + security watch — dennisyu.com & blitzmetrics.com | Mondays 11:01 | Redirect-Hygiene Auditor | Crawl health, redirect chains, Search Console issues, malware and hacked-page checks. |
| GitLab access — chase until site-builder token lands | Daily 14:02 | Stephanie — Operations | An access-blocker chaser. Retires itself when the token arrives. |
| Scheduled-jobs health audit | Weekly, local | Fleet Uptime Monitor | Replaced the old cloud Weekly Fleet Audit on 7/28/26. Probes the BlitzAdmin/LSS fleet for 5xx and drafts — never sends — an alert on a fleet-wide failure. |
The factory floor, stage by stage
Prerequisite · Plumbing
| Agent | What it owns | Skill file |
|---|---|---|
| Christopher — Local Service Website Auditor ● | Audits a local service website against the Quick Audit checklist and the five digital-plumbing checks: GBP, location pages, call tracking, pixels, CRM data exchange. | in the article |
Control plane · Access & Governance
| Agent | What it owns | Skill file |
|---|---|---|
| The RACI Enforcer ● | Dennis is never R or A. Every project gets a young adult as owner; Dennis sits in Consulted and Informed. Enforces it at setup and repairs projects that already drifted. | 060-project-ownership-raci.skill.md |
| The Publisher ● | Every skill has a webpage. Enforces the three artifacts — skill file, definitive article, roster entry — in the same pass a skill ships. | 061-skill-publishing-standard.skill.md |
| Fleet Uptime Monitor ● | Probes the BlitzAdmin/LSS fleet for HTTP 5xx outages and drafts — never sends — an alert on a fleet-wide failure. Runs on the weekly schedule above. | 062-fleet-uptime-monitor.skill.md live |
| Stephanie — Operations Assistant (coming soon) | Access, asset tracking, onboarding, and the governance discipline that keeps the factory shipping. | — |
Stage 1 · Produce
| Agent | What it owns | Skill file |
|---|---|---|
| Ethan — Positive Mentions ● | Finds, verifies, and scores every good thing the market already says about you on a 30-point authority scale, into a ranked proof library. | 010-positive-mentions-harvester.skill.md |
| Company KB Builder ● | Builds the structured company knowledge base that every downstream stage reasons from. | in the article |
| Olivia — Knowledge Panel Helper ● | Builds and maintains the entity signals that earn a Google Knowledge Panel. | in the article |
| The Positioning Strategist ● | Names who you really are to the market — buy box, passion, and the one-sentence differentiation only you can claim — by mining your own AI history, bio, and deal record. Run this before the KB Builder. | 011-personal-brand-strategist.skill.md live |
| Capture Coordinator (coming soon) | Coordinates capture of the raw ingredients: job-site photos, one-minute videos, customer love, podcasts, speaking, mentions. | — |
Cross-cutting · Reason / Plan
| Agent | What it owns | Skill file |
|---|---|---|
| Topic Wheel Generator (coming soon) | The bridge. Reasons from the knowledge base about WHO to build relationships with and WHAT to talk about. Pre-condition for everything downstream. | — |
Stage 2 · Process
| Agent | What it owns | Skill file |
|---|---|---|
| Brandon — Blog Post Helper ● | Turns transcripts and raw ingredients into house-style articles. Never originates; always transforms real material. | 002-brandon-blog-post-helper.skill.md |
| Jennifer — Article Grader ● | Grades every article against E-E-A-T and the house standard. Nothing ships below B+. Loops with Brandon until A−. | 001-jennifer-article-grader.skill.md |
| The Video Editor — Descript ● | Edit-by-text processing of raw video and audio: clean filler and pauses, captions, transitions, highlight reels, short clips, publish a share link. Non-destructive. | 003-video-editing-descript.skill.md live |
| Library Curator (coming soon) | Keeps the Content Library organized: pillar → micro → distribution, every asset tagged to the Topic Wheel. | — |
Stage 3 · Post
| Agent | What it owns | Skill file |
|---|---|---|
| Personal Brand Site Engineer ● | Engineers personal brand websites to the BlitzMetrics spec: entity home, fast, complete, crawlable. | in the article |
| The Repurposer ● | Watches a YouTube channel daily, judges which new videos deserve repurposing, and turns winners into published articles on the right site — graded A− by Jennifer, cross-linked per Content Factory hub rules. Real run: 26 of Richard Canfield’s videos. | 004-video-repurposing-agent.skill.md live |
| BlitzAdmin Fleet Manager ● | The publishing plumbing under Stage 3: lists managed sites, publishes content, audits personal brand sites, manages WordPress across the whole BlitzAdmin fleet. | plugin: blitzadmin-wordpress |
| Adrian — Local Service Page Builder (coming soon) | Builds location service pages that link to GBP and load fast. | — |
| Cross-Posting Agent (coming soon) | Distributes every processed asset to every channel that earns authority, formatted per channel. | — |
Stage 4 · Promote
| Agent | What it owns | Skill file |
|---|---|---|
| Redirect-Hygiene Auditor ● | Audits redirects so no earned authority leaks through broken or chained URLs. Runs on the weekly SEO watch above. | in the article |
| The Meta Ad Agent ● | Runs and optimizes Meta lead-gen ads for a single client through the Meta Ads MCP — one skill, many clients, parameterized by a client profile. The execution layer beneath the Boost Team. | 047-facebook-ad-agent.skill.md live |
| Google Ads Agent (coming soon) | Extends Dollar a Day economics to Google Search, Display, and YouTube campaigns. | — |
Feedback · Perform
| Agent | What it owns | Skill file |
|---|---|---|
| The MAA Reporter ● | The primary feedback-loop agent: a weekly Measure-Analyze-Act brief that reports results and feeds Reason / Plan and the next Produce cycle. | 046-maa-reporter-weekly-brief.skill.md |
The Dollar a Day Boost Team
Seven agents, one conductor. They run the entire Dollar a Day master course — the $1/day promotion strategy for Stage 4. The sequence is fixed: the Unicorn Hunter finds it, the Signal Setter frames it, the Campaign Builder launches it, the Boost Scorer judges it, the Funnel Sequencer connects the winners, the MAA Reporter closes the week, and the Boost Manager conducts all of it. The operator only ever talks to the Boost Manager.
| Agent | What it owns | Skill file |
|---|---|---|
| The Boost Manager ● | The conductor. Takes the GCT brief, runs the daily/weekly loop, coordinates the six specialists, enforces kill/keep/scale, ships the weekly report. | 040-boost-manager-dollar-a-day.skill.md |
| The Unicorn Hunter ● | Reads organic performance, finds the proven winners, never boosts cold content, assembles the 3×3 candidate grid. | 042-unicorn-hunter-content-selection.skill.md |
| The Signal Setter ● | Decides per network whether to go broad-and-signal-led or lightly targeted, and writes the three-second hook that tells the algorithm who the content is for. | 041-signal-setter-audience.skill.md |
| The Campaign Builder ● | Builds the $1/day boost, the 7-day layering schedule, objective, placements, pixel/event, and naming. | 043-campaign-builder-boost-mechanics.skill.md |
| The Boost Scorer ● | Runs the daily kill/keep/scale decision against benchmarks, watches the 10% saturation ceiling, switch-boosts winners to new audiences. | 045-boost-scorer-optimizer.skill.md |
| The Funnel Sequencer ● | Turns winners into the WHY → HOW → WHAT retargeting sequence. Nine pieces, one funnel. | 044-funnel-sequencer-3x3.skill.md |
| The MAA Reporter ● | The loop-closer. The weekly Measure-Analyze-Act brief that reports results and feeds the next Reason / Plan and Produce cycle. | 046-maa-reporter-weekly-brief.skill.md |
Vertical spin-offs
The same Dollar a Day machine, tuned per trade — each with its own definitive article and skill file.
| Agent | What it owns | Skill file |
|---|---|---|
| Dollar a Day for Landscapers ● | Hooks, benchmarks, and seasonal content patterns for the trade. | 050-dollar-a-day-landscapers.skill.md |
| Dollar a Day for HVAC ● | Emergency-call hooks, seasonal demand, and trade benchmarks. | 051-dollar-a-day-hvac.skill.md |
| Dollar a Day for Dentists ● | Patient-story content and local trust signals. | 052-dollar-a-day-dentists.skill.md |
| Dollar a Day for Real Estate ● | Listing content, neighborhood authority, and geo signals. | 053-dollar-a-day-real-estate.skill.md |
| Dollar a Day for Dunkers ● | Tuned for basketball dunkers and athlete personal brands. | 054-dollar-a-day-dunker.skill.md |
| Dollar a Day for Roofers ● | Storm-response content and before/after proof. | 055-dollar-a-day-roofers.skill.md |
| Dollar a Day for Pest Control (coming soon) | Built; publishes when the target site’s REST route is enabled. | 056-dollar-a-day-pest-control.skill.md |
| Dollar a Day for Law Firms (coming soon) | Built; publishes when the target site’s REST route is enabled. | 057-dollar-a-day-law-firms.skill.md |
The Shared Memory crew
Install the cross-agent shared-memory protocol in the shared Obsidian/Markdown vault; Dennis OS shows the complete arrangement.
Added July 2026. Two agents that give every model on the team one shared brain instead of each one starting cold. How they came to exist — two AIs colliding in one folder — is documented in the shared-brain meta-article; build your own from the definitive article.
Boots every model from one shared brain and signs its work — vendor, exact model, decisions, next steps. Built by Claude, QA’d by Codex.
Searches, validates, secret-scans, and packs the shared memory for any AI. Built by Codex, adopted by Claude in the merge.
Utility skills every agent leans on
Not agents — these are the shared hands. Any agent above can call them, and they have no client-facing article because they produce artifacts rather than outcomes.
| Skill | What it does |
|---|---|
docx / pptx / xlsx / pdf |
Build and read Word documents, decks, spreadsheets, and PDFs. Every client deliverable that is not a webpage goes through one of these. |
dataviz |
The house chart standard. Read before writing a single line of chart code so every visualization across every client reads as one system. |
skill-creator |
Creates, edits, and evaluates skills — including running evals and tuning a description for trigger accuracy. Pairs with The Publisher: skill-creator builds it, The Publisher ships the page. |
claude-in-chrome |
The hands for anything with no API: Basecamp, self-hosted WordPress, Knowledge Panel work, any admin UI. |
The gap — closed August 3, 2026
This section used to list six skills that were built and running with no canonical page — a direct violation of The Publisher’s own rule. All six shipped on August 3, 2026:
| Article | Stage | What it closed |
|---|---|---|
| The Repurposer | Stage 3 · Post | The daily channel watchdog, validated on 247 Escape Fitness episodes. |
| The Meta Ad Agent | Stage 4 · Promote | The execution layer under the Boost Team. |
| The Video Editor | Stage 2 · Process | Fills the long-empty Descript slot. |
| The Positioning Strategist | Stage 1 · Produce | The step-1 brief every other skill aims at. |
| Fleet Uptime Monitor | Control plane | Two real incidents documented, including the 111-day-stale roster. |
| BlitzAdmin Fleet Manager | Stage 3 · Post | The publishing plumbing everything else stands on. |
What remains is not a documentation gap. Dollar a Day for Pest Control and for Law Firms are built and waiting on the target sites’ REST routes — infrastructure, not writing. The remaining coming soon agents have a fixed place in the factory but no skill file yet; they are a build queue, not a publishing queue.
For AI agents reading this page
The machine-readable registry is embedded as <script type="application/json" id="agent-registry">. Parse object_types, operating_model, production_phases, roster_buckets, and each record’s operating_bucket; do not infer that a skill variant is an agent or that a team is a phase. To attempt a published agent, fetch its definitive_article, obtain the referenced skill file, configure only the declared connectors, honor the permission tier, and verify the task’s acceptance criteria. Published means instructions are available, not that E1–E4 capability has been established. Permission tiers are read, draft, and publish; public or client-visible action requires explicit human approval per action.