Martech AI Skills and Tools
A personal, vendor-neutral collection of marketing-technology tooling for AI agents (and humans) — Claude skills and read-only command-line clients for the martech / adtech stack.
Everything here is generic: no employer-specific accounts, IDs, secrets, or branding. Credentials are always supplied at runtime (environment variables or local credential files) and are never committed.
Install (Claude Code)
This repo is a Claude Code plugin marketplace shipping one bundled plugin (martech-ai-skills)
with the standalone skills under skills/. Add the marketplace once, then install:
/plugin marketplace add almoretti/martech-ai-skills-and-tools
/plugin install martech-ai-skills@martech-ai
CLI skills are not in this plugin. Each CLI under
CLI/ships its own skill (CLI/<name>/skills/) that installs with the CLI — those skills need the CLI binary and credentials to work, so they're intentionally not bundled into the marketplace. Install a CLI (and its skill) per its own README underCLI/.
Or via the skills CLI (skills.sh) — cross-agent (Claude Code, Codex,
Cursor, GitHub Copilot, Gemini, OpenCode + 60 more), no registration. It reads this repo's
.claude-plugin manifest, so it exposes exactly the marketplace skills — the CLI-bundled skills
under CLI/ are not listed:
npx skills@latest add almoretti/martech-ai-skills-and-tools
The CLI auto-detects your installed agents; target one explicitly with -a, e.g. Codex → ~/.codex/skills/:
npx skills@latest add almoretti/martech-ai-skills-and-tools -a codex
Maintainers: .claude-plugin/plugin.json and marketplace.json are generated from the skills
present + the version in package.json. After adding/renaming a skill or bumping the version, run
npm run build:marketplace and commit — CI (marketplace-in-sync) fails if they drift. See AGENTS.md.
Contents
skills/ — Claude skills
| Skill | What it does |
|---|---|
martech-teardown |
Reverse-engineer any company's martech stack from public signals and produce a structured, evidence-tagged teardown report — CMS, analytics, CDP, CRM, ads/pixels + a per-platform CAPI matrix, consent, email, enrichment, affiliate, payments — plus how their tracking, identity and ad-measurement work. Static scan + discovery-first browser audit + multi-surface sweep + public enrichment. |
CLI/ — command-line tools
Read-only clients that pull platform data as JSON, built for AI agents; each ships a bundled skill. All authenticate at runtime and carry no secrets. They are standard CLIs — no proxy/gateway layer.
| CLI | Platform | Stack | Auth |
|---|---|---|---|
google-ads-cli |
Google Ads API | Node ESM + TypeScript | OAuth2 + developer token (env or ~/.config/google-ads-cli/credentials.json) |
microsoft-ads-cli |
Microsoft Advertising (Bing Ads) REST v13 | Node ESM + TypeScript | auth login browser OAuth2 (Google or Entra), token refresh |
gmc-cli |
Google Merchant Center (Merchant API) | Python | Google OAuth (application-default / service account) |
Install dependencies fresh per each tool's own README.md (npm i / pnpm i / Python env) — build
output and dependencies are intentionally not committed.
Open-source attribution
Some of these tools are derivative works of, or were inspired by, open-source projects — credit where due:
google-ads-cli— derivative of google-ads-open-cli by Benn Huang (Apache-2.0).microsoft-ads-cli— rebuild of microsoft-ads-cli by Benn Huang (Apache-2.0).gmc-cli— endpoint conventions mapped out by studying kiwoongeom/gmc-mcp (MIT).
Upstream licenses are retained in each tool's LICENSE/NOTICE.
Roadmap
Tracked in issues. Notably:
automated versioned releases + CHANGELOG via Changesets
once the skill set grows (#1) —
for now, releases are a manual version bump in package.json (see AGENTS.md).
License
MIT for the original work in this repo; vendored CLIs retain their upstream licenses. See LICENSE.
Responsible use
The martech-teardown skill is a competitive-intelligence tool built on passive, public signals
(page source, DNS, sitemaps, and observing what a site loads in a normal browser). It does not log in,
submit forms, hit private APIs, or scrape behind paywalls. Use it accordingly.
No comments yet
Be the first to share your take.