| name | marketing-md-setup |
| description | Set up or retrofit MARKETING.md (Marketing-as-Code) using the core + companion architecture from the Marketing-MD-Agentic-Standard. Use when the user asks to create MARKETING.md, migrate from ad-hoc marketing prompts/project-context files, or standardize agentic marketing context across tools. |
| compatibility | Markdown + YAML. Works best in repos that use Agent Skills and version-controlled AI context files. |
marketing-md-setup
Practical workflow for adopting MARKETING.md as a machine-readable marketing system of record for AI agents.
Use this skill when a user wants to:
- initialize
MARKETING.md in a new repository;
- migrate from scattered prompt notes or
project-context style files;
- standardize brand voice, personas, campaign defaults, and safety rules for multiple agents;
- adopt the core/companion architecture for role-based context loading;
- enforce conservative guardrails (approval gates, suppression, anti-hallucination) in AI marketing workflows.
Inputs Required
Before creating or editing files, ask and confirm:
- Target scope
- Apply in current repository or another directory?
- Setup mode
from-scratch (new MARKETING.md) or adapt-existing (merge with current docs)?
- Business profile
- What vertical/business model is this for (SaaS, e-commerce, agency, marketplace, etc.)?
- Any jurisdiction/compliance constraints (GDPR, CAN-SPAM, local laws)?
- Agent roles in scope now
- Which agents exist today (content, campaign, paid, lifecycle, analytics, personalization, crisis)?
- Which companion files are required immediately vs later?
- Oversight policy
- Desired autonomy tiers (
full_autonomy, notify_and_proceed, notify_and_wait, human_only)?
- Which actions must always require human approval?
- Change policy
- Merge into existing files (recommended) or replace where safe?
If any answer is ambiguous, stop and ask a follow-up instead of guessing.
What To Do First
- Inventory existing context files.
- Check for
MARKETING.md, project-context*, .agents/, AGENTS.md, .github/skills/**, .codex/**, .claude/**, .cursor/**.
- Detect existing marketing docs.
- Check for strategy docs, campaign playbooks, voice/tone docs, KPI docs, compliance notes.
- Choose setup flow.
- No
MARKETING.md: run from-scratch.
- Existing docs/context: run
adapt-existing and preserve prior intent.
- Establish safety baseline.
SAFETY.MD must be created/updated first among companion files.
Preflight Detection
Run in target repository:
pwd
rg --files -g 'MARKETING.md' -g 'companion/**/*.MD' -g 'companion/**/*.md' -g '*project-context*' -g 'AGENTS.md' -g '.agents/**' -g '.github/skills/**' -g '.codex/**' -g '.claude/**' -g '.cursor/**'
rg -n "brand voice|persona|kpi|approval|suppression|utm|campaign" -g '*.md'
Interpretation:
- If
MARKETING.md is missing, scaffold a core file plus companion directory.
- If
MARKETING.md exists, patch incrementally and avoid destructive rewrites.
- If only
project-context exists, map it into structured MARKETING.md sections.
Canonical File Architecture
Expected structure:
MARKETING.md
companion/
SAFETY.MD
BUDGET.MD
CALENDAR.MD
CAMPAIGNS.MD
CHANNELS.MD
COMPETITORS.MD
CONTENT.MD
EXPERIMENTATION.MD
LIFECYCLE.MD
MEASUREMENT.MD
MESSAGING.MD
PARTNERS.MD
PERSONALIZATION.MD
PRICING.MD
SIGNALS.MD
VISUAL.MD
Loading policy:
- Every agent session must read
MARKETING.md + SAFETY.MD.
- Other companion files load by role and task.
Flow A: From Scratch
Step 1: Create core + companion skeleton
Create:
MARKETING.md
companion/
companion/SAFETY.MD (mandatory first companion)
Add remaining companion files only when required by active agent roles.
Step 2: Add minimum viable MARKETING.md sections
At minimum include:
- What this file is (authoritative agent config)
- Business identity
- Voice & tone
- Customer personas
- Campaign types and defaults
- KPI framework
- Suppression & compliance rules
- Human approval gates
- Prohibited actions
- Versioning and review cadence
- AI autonomy/safety guardrails
- Companion file registry
Use YAML for machine-readable policy and prose for qualitative guidance.
Starter skeleton:
# MARKETING.MD
## What This File Is
Core marketing configuration read by all agents at session start.
## 1. Business Identity
```yaml
brand:
name: "[Fill in]"
category: "[Fill in]"
business_model: "[Fill in]"
2. Voice & Tone
- Core voice attributes
- Channel-specific tone modulation
- Allowed/prohibited language
3. Customer Personas
- Persona definitions
- Persona quick reference table
4. Campaign Types
- Campaign type defaults and KPI mapping
5. KPI Framework
anomaly_detection:
lookback_window_days: [N]
sensitivity: "[low | medium | high]"
6. Suppression & Compliance Rules
global_suppression:
unsubscribes: immediate_permanent
7. Human Approval Gates
- Define
notify_and_wait decisions explicitly
8. Prohibited Actions
- Never send to suppressed addresses
- Never launch paid without human confirmation
9. Versioning
marketing_md:
version: "1.0.0"
review_cadence: "Quarterly"
10. AI Autonomy & Safety Guardrails
autonomy_levels:
full_autonomy: []
notify_and_proceed: []
notify_and_wait: []
human_only: []
11. Companion File Registry
- Registry with role-based loading rules
### Step 3: Create SAFETY.MD immediately
`SAFETY.MD` should include:
- crisis severity levels;
- kill switch conditions and owner;
- escalation matrix;
- holding statement templates;
- blocked topics/claims list;
- restart criteria after incident.
### Step 4: Add role-driven companions
Add only files needed for current agent roles.
Examples:
- Content agent: `MESSAGING.MD`, `CONTENT.MD`.
- Campaign/Paid agent: `CAMPAIGNS.MD`, `BUDGET.MD`, `PRICING.MD`, `CHANNELS.MD`.
- Analytics/Optimization agent: `MEASUREMENT.MD`, `EXPERIMENTATION.MD`.
- Lifecycle/Triggered agent: `LIFECYCLE.MD`, `SIGNALS.MD`, `PERSONALIZATION.MD`.
## Flow B: Adapt Existing Project
### Step 1: Map existing docs to MARKETING.md sections
Typical mapping:
- brand deck / positioning notes -> Business Identity + Voice/Tone
- ICP docs -> Personas
- campaign playbooks -> Campaign Types
- dashboard/KPI docs -> KPI Framework
- legal/compliance docs -> Suppression & Compliance
- internal approval SOP -> Human Approval Gates
### Step 2: Merge, do not overwrite
When files already exist:
- preserve IDs and naming conventions where possible;
- keep valid existing decisions unless user requests replacement;
- convert unstructured rules to YAML blocks when they control behavior.
### Step 3: Enforce anti-hallucination structure
Add or validate in `MESSAGING.MD`:
- approved claims library (human-curated);
- explicit rule: omit unverified claims;
- source requirement for stats/testimonials/comparative claims.
### Step 4: Encode conservative fail defaults
Ensure fallback behavior is explicit:
- unknown suppression status -> treat as suppressed;
- unknown budget status -> pause spend and alert;
- missing channel rule -> do not execute action;
- ambiguous compliance state -> escalate to human.
## Validation Checklist
Before completion, confirm:
- `MARKETING.md` exists and includes core policy sections.
- `companion/SAFETY.MD` exists and is referenced as universal load.
- Autonomy tiers are explicit and actionable.
- Approval gates and prohibited actions are machine-readable or unambiguous.
- Claims governance exists (`approved claims` policy).
- Review cadence is defined (recommended: quarterly).
- Files are in version control and update history is trackable.
## Troubleshooting
1. File is too long/noisy for every agent session
- Keep only universal rules in `MARKETING.md`.
- Move role-specific detail to companions.
2. Agents produce inconsistent messaging
- Tighten personas and channel tone rules.
- Centralize facts in `MESSAGING.MD` approved claims library.
3. Unsafe autonomous behavior
- Move risky actions into `notify_and_wait` or `human_only`.
- Add circuit breakers in `SAFETY.MD`.
4. Existing docs conflict with new structure
- Preserve old docs as source material, but create canonical mappings in `MARKETING.md`.
- Ask user which source is authoritative where conflicts remain.
## Execution Guidance
When invoked, do this sequence:
1. Run preflight inventory.
2. Confirm setup mode (`from-scratch` or `adapt-existing`).
3. Build/update `MARKETING.md` core.
4. Ensure `SAFETY.MD` exists and loads universally.
5. Add only necessary companion files for active roles.
6. Validate autonomy, approvals, suppression, and claims governance.
7. Summarize file tree and recommended next companion files.
## References
- https://github.com/benkohcc/Marketing-MD-Agentic-Standard
- https://raw.githubusercontent.com/benkohcc/Marketing-MD-Agentic-Standard/main/MARKETING.md
- https://ddpa.ru/kb/ai/marketing/
- https://ddpa.ru/kb/ai/marketing/marketing-md/
- https://ddpa.ru/kb/ai/marketing/marketing-md-standard/
- https://ddpa.ru/kb/ai/marketing/marketing-skills/