mcpField
M
media-buyer

media-buyer

github.com/kelpi-aiMarketing & growth559 installsUpdated 17w ago
Repository

About media-buyer

A marketing skill for building small-budget Meta Ads test campaigns from approved creative angles. It creates a paused, reviewable campaign structure and checks the conversion event before anything can be published.

Key features

  1. 01Single-campaign test structure
  2. 02Broad targeting with one ad set
  3. 03$20/day starting budget
  4. 04Ads mapped to approved angles
  5. 05Conversion-event and pixel check
  6. 06Paused drafts pending review

Use cases

  1. 01Setting up a low-budget angle test
  2. 02Preparing ads for human review before publishing
  3. 03Checking conversion tracking before launch

MCP support

Requires a Meta Ads MCP with write access to build campaigns. Campaigns remain paused until a human approves publication.

Install

Copy the command and run it in a terminal.

$npx skills add https://github.com/kelpi-ai/meta-ads-skills --skill media-buyer

After installing, type/media-buyerin Claude Code to confirm it loaded.

Bundled filesskill.meta.json

Details

From the project README

Media Buyer

Doctrine

Simple structure wins at small budgets: one campaign, one broad ad set, 3-5 genuinely different angles as separate ads. Fragmenting budget across many ad sets starves the algorithm of signal; low budgets get killed by too much structure, not too little. Targeting stays broad on purpose: post-Andromeda, the ad content is the targeting, and the $20/day test exists to find the angle that closes the loop before real money goes in.

When to use

  • After Creative Director; you have approved ads and image hashes.
  • Needs a Meta Ads MCP with write access. Everything is built paused; nothing spends until you publish.

Run it

Build me a test campaign, everything PAUSED until I review it:

1. Campaign: sales objective (or leads, if my offer is lead-gen), one campaign only.
2. Ad set: one, broad targeting (country + broad age range only, no interest stacks), $20/day budget.
3. Ads: one ad per approved angle, using these creatives: [IMAGE HASHES + COPY]. Do not blend angles.
4. Confirm the pixel/conversion event it optimizes toward, and flag if the pixel looks dead or misconfigured BEFORE anything else.
5. Show me the full structure as a table (campaign > ad set > ads with their angles) and wait. I will say "publish" when I have reviewed it.

Do not enable any Advantage+ creative enhancements. I want the creatives to run exactly as approved.

Guardrails

  • Paused draft first, always. The human says "publish", the skill never assumes it.
  • $20/day until an angle has proven itself with 2-3x target cost-per-result of spend. Scaling before signal is donating to Meta.
  • One campaign. If you feel the urge to add a second ad set, you probably need a different angle instead.
  • After publishing: hands off for a week. Daily edits reset learning. Reading daily is what the Daily Auditor is for.

Good output looks like

A paused campaign whose structure you can read in ten seconds, optimizing toward a verified conversion event, with each ad mapped to a named angle.

Comments

Markdown supported