# Marketing Os > Marketing OS - Skill: `everyinc/marketing-os` (Agent Skill, multi-file: 14 files) - Install (CLI): `npx skillmds@latest add everyinc/marketing-os` - Raw SKILL.md: https://api.skillmd.com/api/skills/everyinc/marketing-os/raw - Safety review: pending - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: Marketing & Growth - Author: everyinc (https://skillmd.com/u/everyinc) - Updated: 2026-09-17 - Page: https://skillmd.com/skills/everyinc/marketing-os --- # Marketing OS Every's in-house marketing operating system, authored by Douglas Brundage as Head of Marketing. This is the theoretical and methodological foundation that every downstream skill in the repo — brand-voice, positioning, craft, strategy, launches — pulls from at invocation time. The goal: an AI collaborator that is first a master marketer grounded in the discipline's best thinking, then Every-specific as brand docs and frameworks come online. Marketing OS ensures that no skill operates in a vacuum. Every piece of generated, edited, or evaluated work inherits a shared point of view, a shared canon, and a shared set of operational frameworks. ## System Marketing OS is composed of eight reference files, each serving a distinct role: ### references/canon.md The reference encyclopedia. Thinkers, practitioners, and researchers organized by domain (effectiveness, creative, brand theory, behavioral, strategy, deep history). Load this when a task needs historical grounding, a specific thinker's lens, or evidence-based reasoning from the marketing canon. ### references/method.md The opinionated layer. Every's argumentative positions on effectiveness, creative, brand theory, behavioral science, taste, AI-native marketing, and what makes Every different. Load this when a task needs the point of view — when the work reflects what we believe rather than only what the field knows. ### references/frameworks.md The operational reference. Battle-tested frameworks (Binet/Field, Sharp, Keller, Aaker, Romaniuk, Rumelt, Thaler/Sunstein, Christensen, Trout/Ries) with deployment guidance. Load this when a task needs to operationalize a framework — when we need to apply a model in context. ### references/collaboration.md How AI collaborators work with Douglas: interview before ambiguous long-form work, preserve creative specificity, use active language, and learn from his edits. Load this when developing or revising substantial strategy, positioning, campaign, or brand work. ### references/native-document-editing.md How to edit existing native documents without destroying their structure or design. Load this before changing a formatted, multi-tab, table-heavy, or collaborative document. ### references/knowledge-boundaries.md How project work becomes reusable operating doctrine without contaminating future work with dated briefs, campaign facts, or one-off preferences. Load this when adding, revising, or compounding a skill. ### references/learning-loop.md How normal work compounds: detect meaningful feedback or results, capture the smallest useful learning, classify it, and promote only evidence that clears the knowledge boundary. Load this after a meaningful correction, explicit ruling, measured result, or repeated pattern changes a future decision. ### references/golden-path.md The canonical Compound Marketing workflow. Load this for substantial strategy, recurring marketing, or GTM work that should carry governed decisions from context through delivery and learning. ## When to invoke This skill is loaded automatically by every other skill in the repo. It is the root dependency. Specific situations where it is invoked directly: - When generating any brand or marketing work (copy, positioning, strategy, launches) - When editing or evaluating existing work against Every's standards - When building or auditing a new skill for alignment with the OS - When onboarding a new brand or product into the Every system - When resolving a conflict between brand-voice and positioning guidance ## Reference routing When this skill is loaded, route to the appropriate reference file based on the task: | Task type | Load | |---|---| | Need a reference, thinker, or historical grounding | `references/canon.md` | | Need the point of view or argumentative position | `references/method.md` | | Need to operationalize a framework (Binet/Field, Sharp, Aaker, etc.) | `references/frameworks.md` | | Need to develop or revise substantial work with Douglas | `references/collaboration.md` | | Need to edit an existing formatted document | `references/native-document-editing.md` | | Need to add or compound repository guidance | `references/knowledge-boundaries.md` | | Meaningful feedback or a result should improve future work | `references/learning-loop.md` | | Substantial strategy, recurring marketing, or GTM work | `references/golden-path.md` | | 10- to 15-minute GTM approval deck for a launch content ecosystem | `../../marketing/gtm/SKILL.md` and `../../marketing/gtm/references/content-system-approval-deck.md` | Multiple references can be loaded simultaneously. A positioning task, for example, might load canon.md for Trout/Ries context, method.md for Every's positioning stance, and frameworks.md for the operational positioning template. ## Close the learning loop Do not add a questionnaire or ceremonial retrospective. When normal feedback, an explicit ruling, a measured result, or a repeated pattern supplies evidence that should change a future decision, load `references/learning-loop.md` and capture the candidate as part of the task closeout. If the evidence teaches nothing transferable, finish the task without creating a learning record. Never promote a candidate merely because it is memorable or recent. Apply the knowledge boundary, protect sensitive project detail, and update the narrowest source only when the promotion standard is met. After changing root doctrine, routing, collaboration, knowledge boundaries, or this learning loop, run the five cases in `evals/smoke-cases.json` against `evals/smoke-rubric.md`. Use `scripts/check_learning_loop.py` to validate the case manifest and any learning candidate before it enters the repository. ## Dependencies None — this is the root dependency. All other skills load this skill. ## Interaction with downstream skills - **brand-voice/** skills inherit the method's creative and taste stances - **positioning/** skills inherit the method's brand theory weighting and the frameworks - **craft/** skills inherit the canon's creative lineage and the method's creative stack - **strategy/** skills inherit the full OS — canon, method, and frameworks - **launches/** skills orchestrate all of the above through the OS - **marketing-science/** skills contribute back to the OS — research and archetyping inform the canon and frameworks over time - **measurement/** connects Marketing-created demand with subscriber quality, brand movement, authority, and business outcomes - **compound-from-results** turns completed programs into decisions, program guidance, and narrowly supported updates to durable doctrine - **strategy/compound-marketing** carries governed decisions through strategy, market execution, and a measured learning closeout - **marketing/gtm** is the single GTM entry point across flagship, standard, revision, drumbeat, and social modes - **craft/compound-copywriting** turns accepted edits into bounded project guidance and tests transfer on fresh copy before any durable promotion - **strategy/** preserves the full idea while producing concise, decision-ready briefs - **strategy/marketing-pod** coordinates specialist skills across parallel agent roles while Douglas retains strategic decisions, approval, and final taste authority ## Knowledge boundary Project work supplies evidence. It does not become operating doctrine automatically. Reusable skills contain transferable methods; brand and product modules contain durable ruled canon; optional historical cases live under `docs/cases/` and are never loaded by default. Apply `references/knowledge-boundaries.md` before any repository update. ## Status The doctrine, collaboration rules, unified GTM route, and Compound Marketing run records are active. Compound Copywriting and marketing pods remain beta until prospective use clears their stated transfer and review-burden thresholds. Files marked `## To do` are scaffolding and never route as authoritative guidance.