AI Marketing Skills
Twenty marketing frameworks packaged as SKILL.md files that Claude Code and other AI assistants can actually execute.
- Source repo
- BrianRWagner/ai-marketing-claude-code-skills
- Stars
- ★ 459
- Last updated
- 6d ago
- License
- MIT
- Primary language
- JavaScript
- FA score
- 46/100 · Major gaps
At a glance
- Works with
- Portable with changesCodex · Claude CodeChatGPT · Claude.ai (Partial support)
- You'll need
- Typical use
- A founder or solo marketer who wants a structured positioning review without hiring a consultant runs positioning-basics and brand-positioning-audit for a scored diagnosis.
- Main limitation
- There is no standalone agent in the repository; results depend entirely on the underlying assistant and can vary by model.
- Website
- brianrwagner.com ↗ set on GitHub
- Maintenance
- Commits in the last 30 dayslast commit 2026-11-14+1.0 ★/day over 7 days
- Source review
- 46/100 · Major gaps
What does this agent do, and when should you use it?
This repository is a library of 20 Markdown SKILL.md files that teach an AI assistant how to run specific marketing workflows: positioning, brand voice, content ideation, competitor teardowns, conversion audits, and daily operations. It is not a standalone executable; the skills are written briefs that an agent loads and follows, which makes Claude Code the native target while Codex, ChatGPT, Cursor, and Grok can use them by pasting or uploading the SKILL.md. Supporting Bash scripts (scripts/install.sh, scripts/list-skills.sh, scripts/convert.sh) detect installed platforms and convert files into formats such as Cursor .mdc rules. As of v4.1 every skill supports quick, standard, and deep execution modes, and outputs are text artifacts like scored audits, prioritized fix lists, voice specs, and rewritten copy. The project is MIT licensed and free, but it needs an LLM assistant that can read the skill files.
Each skill is a SKILL.md that defines trigger phrases, a questioning framework, analysis steps, and an output format. Once the folder is placed in ~/.claude/skills/ or a platform rule directory, mentioning a related topic makes the assistant follow that procedure: positioning-basics walks through audience, offer, and differentiation; ai-discoverability-audit scores how ChatGPT, Perplexity, Claude, and Gemini describe a brand, lists cited sources, and produces a prioritized fix list with a worksheet and prep script; brand-positioning-audit returns a scored diagnostic with specific fixes; competitor-intel-brief tears down one named competitor across positioning, ICP, pricing signals, channel patterns, moat, and vulnerabilities; voice-extractor and brand-voice-extractor derive a voice guide from writing samples (the latter can also build a voice from scratch); cold-outreach-sequence builds LinkedIn and email outreach sequences; homepage-audit reviews a landing page for conversion issues; tweet-draft-reviewer scores drafts 1–10 against 8 voice rules and rewrites anything below 7; vault-cleanup-auditor scans an Obsidian vault for stale drafts, empty folders, and duplicate filenames and saves a dated report. scripts/install.sh auto-detects Claude Code, Cursor, and Windsurf, while scripts/convert.sh emits .mdc or rule-frontmatter .md files.
- A founder or solo marketer who wants a structured positioning review without hiring a consultant runs positioning-basics and brand-positioning-audit for a scored diagnosis.
- A content creator using Claude Code to plan posts uses content-idea-generator and linkedin-authority-builder to define pillars, formats, and a posting rhythm.
- A B2B seller preparing for a call uses meeting-prep to get participant research, context, an agenda, and sharp questions.
- A marketing team checking AI search visibility runs ai-discoverability-audit to see which sources models cite and what to fix first.
- A writer who wants AI to sound like them uses voice-extractor or brand-voice-extractor to produce a reusable voice spec from samples.
- An Obsidian user tidying their vault runs vault-cleanup-auditor to find stale drafts and duplicates and get a dated report.
Key setup facts
Pulled from the install and usage notes below. The project README has the current values.
- Config files
~/.claude/skills
How do you install or deploy this agent?
You need Git, Bash, and an AI assistant that can read skill files (Claude Code is the native target). Manual install:
git clone https://github.com/BrianRWagner/ai-marketing-claude-code-skills.git
mkdir -p ~/.claude/skills
cp -r ai-marketing-claude-code-skills/* ~/.claude/skills/Or use the bundled installer, which auto-detects platforms:
cd ai-marketing-claude-code-skills
bash scripts/install.shThe installer accepts flags:
bash scripts/install.sh --all # install to every detected platform
bash scripts/install.sh --platform=cursor # specific platform only
bash scripts/install.sh --dry-run # preview without writing filesCursor and Windsurf use project-level rule directories, so convert the skills first:
bash scripts/convert.sh --platform=cursor --output-dir=./.cursor/rules
bash scripts/convert.sh --all # convert for all platforms into ./converted/The repository documents no API keys; the actual reasoning comes from whatever AI assistant you use.
How do you use this agent?
After installing, open Claude Code and describe the task (for example "help me with positioning" or "how do we beat X?"); the matching skill is picked up automatically. To see what is installed:
bash scripts/list-skills.shOn non-native platforms, paste or upload the SKILL.md from the relevant skill folder as the assistant's instructions, for example into a ChatGPT Project or custom GPT. Specify quick, standard, or deep in the prompt to choose execution depth. Outputs are text artifacts — scores, diagnostics, rewrites, teardowns, or dated audit reports — and saving them is up to you.
What are this agent's strengths and limitations?
- All 20 skills are plain Markdown with no runtime code, so any assistant that can follow a written brief can run them and switching tools is cheap.
- Coverage spans the whole marketing funnel — positioning, brand voice, content production, competitor teardowns, cold outreach, landing-page audits — rather than one narrow tool.
- Each skill prescribes an output structure (scores, prioritized fixes, voice specs), which yields more consistent results than free-form prompting.
- Bundled scripts (install.sh, convert.sh, list-skills.sh) auto-detect Claude Code, Cursor, and Windsurf and convert to their rule formats.
- MIT licensed and, as of v4.0, previously paid skills such as the AI Discoverability Audit and Competitor Intel Brief are included free.
- There is no standalone agent in the repository; results depend entirely on the underlying assistant and can vary by model.
- ChatGPT, Claude.ai, Meta Muse, and Grok are manual-only: you must paste or upload SKILL.md, with no official install path.
- Skill quality is not automatically verifiable — the README offers no tests, benchmarks, or sample outputs, so users must judge results themselves.
- Some skills (ai-discoverability-audit, competitor-intel-brief) rely on external research data; the repo provides no scraping or data-collection capability.
How does this agent compare with similar options?
The README names no specific competitors; it distinguishes itself from guides and courses by packaging frameworks as repeatable agent instructions, so the practical alternative is a static marketing tutorial or writing prompts from scratch.
Key facts side by side with the most closely related agents.
| Agent | Source review | Form / cost | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|---|
| AI Marketing Skills This agent | 46 · Major gaps | — | ★ 459 | 6d ago | JavaScript | Codex · Claude Code |
| ASO & App Marketing Skills | 54 · Major gaps | Agent plugin / skillFreemium | ★ 2.3k | 8d ago | MDX | Claude Code |
| AIWriteX Content Studio | 44 · Major gaps | Desktop appFree + model costs | ★ 2.3k | 25d ago | Python | — |
| SuperSEO Skills | 57 · Major gaps | Agent plugin / skillFreemium | ★ 415 | 2mo ago | — | Claude Code · Claude.ai |
How does FollowAgents rate this agent?
Why each dimension lost points
Skills are plain Markdown instructions with no malicious code or credential theft observed. However, install.sh and convert.sh are not in the sample, so their privilege scope cannot be verified (least_privilege 1). Skills read user notes, Obsidian vaults, and website content and write to audits/, voice/, etc., but there is no confirmation step before writes and no data-flow description (user_confirmation, data_flow_transparency, sensitive_data_handling all 1). Dependencies are only referenced scripts/*.mjs and external platforms with no manifest or security notes (dependency_security 1). External effects (fetching sites, publishing content) have no rollback or undo guidance (external_effects, rollback 1). Author attribution and MIT copyright are consistent across headers and LICENSE, so source_attribution scores 2.
SKILL.md files share a similar structure but fields are inconsistent (some have platform, some don't; versions range 1.1.0 to 3.0.0), a self-consistency gap (self_consistency 1). Referenced scripts (audit-prep.mjs, competitive-scan.mjs, lens-router) are absent from the sample, so availability is unconfirmed (dependency_availability 1). Failure handling is limited to scattered gotchas and gate checks with no unified error-message convention (failure_messages 1).
README clearly lists Claude Code native plus Cursor/Windsurf/Codex/ChatGPT/Grok compatibility and manual paths, covering strategy, content, research, conversion, and ops scenarios (audience_and_scenarios 2). Each skill has NOT-for boundaries and quick/standard/deep modes, giving clear capability boundaries (capability_boundaries 2). Descriptions include trigger phrases and exclusions, so trigger precision is decent (trigger_precision 2). Environment fit is documented via a platform table and install locations (environment_fit 2).
README organizes 20 skills by category with links to each SKILL.md, a clear information architecture (information_architecture 2). Install notes include clone, install.sh, flags, and a platform table (install_notes 2). Naming is inconsistent: slugs carry a brw- prefix while names do not, and versions vary across skills (naming_stability 1). Examples are only mentioned in changelogs (references/examples.md) and no actual examples or FAQ appear in the sample (examples_and_faq 1). Known limitations are only scattered gotchas with no consolidated section (1). LICENSE is a full MIT text (license 3). CHANGELOG is referenced but absent from the sample and versions are mixed (versioning_changelog 1). Maintenance responsibility points only to the individual author and a PR process with no governance or SLA (maintenance_responsibility 1).
Skills specify output formats (scorecards, 30-day plans, three case-study formats) that are directly usable (output_usability 2). But they are essentially prompt templates with limited marginal value over asking an LLM directly, and some skills (e.g., daily-briefing with weather) rely on unspecified external data (marginal_value 1). On cost-benefit, install scripts and multi-platform conversion add complexity while benefits are unquantified (cost_benefit 1).
README claims such as 'AI actually executes' and 'AI traffic converts better' have no data source or citation (claim_traceability 1). Cross-file corroboration is limited to consistent author attribution and MIT license; other claims lack cross-source support (cross_source_corroboration 1). Skills distinguish facts from advice (e.g., Tier system), but README marketing claims are not clearly separated from skill content (fact_inference_separation 1).
- Install scripts install.sh and convert.sh are not in the sample; their write scope and privileges cannot be verified, so review them before installing.
- Skills read user notes, Obsidian vaults, and website content and write to local paths without a pre-write confirmation or data-flow description; be cautious with sensitive material.
- README claims such as 'AI traffic converts better' and 'AI actually executes' have no data source and are marketing statements, not facts.
- Referenced scripts/*.mjs and references/examples.md are absent from the sample, so actual availability is unknown.
- Version numbers are inconsistent across skills (1.1.0 to 3.0.0) and CHANGELOG is not in the sample, leaving the upgrade path unclear.
- Publisher identity is not verified by the FollowAgents enterprise registry; this is unknown rather than suspicious, but reliability should not be inferred from it.
FAQ
Are all the skills free?
Do I have to use Claude Code?
Does it need internet access or API keys?
What do quick, standard, and deep modes change?
Does the AI Discoverability Audit query ChatGPT automatically?
Is AI Marketing Skills open source?
Is AI Marketing Skills safe to use?
Is AI Marketing Skills still maintained?
Who makes AI Marketing Skills?
What is AI Marketing Skills's official website?
Thanks — we use this to improve the page.