7 Marketing Skills for GPT-5.6 Sol

GPT-5.6 Sol shipped recently with two things marketers should care about: a skills system, and agents that run in parallel by default.

Most of the coverage is about price and benchmarks. Here is the more useful question for a marketing team: what do you actually point it at?

Because most of what a marketing week needs is not a frontier problem. It is grunt work. The report due Friday. The tracking nobody checked. The week of content that has to ship. The budget question with no clean answer.

So I built skills for exactly that.

7 marketing skills for GPT-5.6 Sol: content, tracking, attribution, reporting, ad waste, campaign briefs and budget. Plus one that runs them in parallel and hands back a single ranked readout.

Free. Open source. Built to do the boring work well so your judgment goes where it matters.

Get all 7 marketing skills for Sol on GitHub

What is a Sol skill?

GPT-5.6 Sol reads skills built on the Agent Skills standard. A skill is a folder with a SKILL.md: YAML frontmatter that tells the agent when to activate, and a Markdown body that defines the job, required inputs, decision rules, output format and guardrails.

That structure matters, because generic prompting produces generic output. A prompt that says "write my report" gives you a template. A skill that knows which metrics matter, what a miss looks like, and how a board reads a report gives you a draft you can send.

Every skill in this pack is data-first. It works from what you paste in: your numbers, your export, your brief. If the data is missing, the skill says so instead of guessing. None of them touch your accounts or change anything live.

The 7 skills

• One Idea Into A Week Of Content - Turns one angle into a full week of channel-mapped drafts, one distinct hook per slot.
• Tracking Signal Health Check - Reviews GTM, GA4 and pixel evidence to flag broken, duplicate or missing tracking before you trust a report.
• Which Channels Actually Drive Revenue - Reads your spend and revenue table, blends the numbers, and surfaces the attribution traps.
• Monthly Marketing Report Builder - Turns the month's numbers and goals into a board-ready report that leads with a verdict.
• Ad Budget Leak Finder - Reviews a campaign or search-term export to find the spend that returns nothing, ranked by money at stake.
• Goal Into A Campaign Brief - Turns a goal and constraints into an execution-ready brief: audience, offer, channels, messaging, budget, measurement.
• Where To Move Next Quarter's Budget - Builds defensible reallocation scenarios from your performance data, with every assumption stated.

The one that uses Sol's parallel agents

There is an eighth skill: Parallel Marketing Sweep.

Hand it everything at once, your export, your channel table, the month's numbers, and it routes each input to the right skill and runs them as parallel subagents on Sol's ultra effort. Instead of running seven skills by hand, you get one reconciled, ranked readout in roughly the time of the slowest single skill.

It only runs the skills whose data you actually gave it. The rest are listed as "not run: missing data", not faked.

Install for Codex

Skills live in ~/.agents/skills for personal use, or in .agents/skills at the root of a repo.

git clone https://github.com/mardab96/codex-sol-marketing-skills.git
mkdir -p ~/.agents/skills
cd codex-sol-marketing-skills
cp -r ad-budget-leak-finder budget-reallocation-planner channel-revenue-reality-check \
  goal-to-campaign-brief idea-to-content-week monthly-marketing-report-builder \
  parallel-marketing-sweep tracking-signal-health-check ~/.agents/skills/

Open Codex and describe the job in normal language; it selects the matching skill. Because these follow the Agent Skills standard, they also run in any other tool that reads that standard.

How to use

Describe the job and hand over the input.

"Here is last month by channel: spend, conversions, revenue. Meta and Google both claim the sale. Where is the money really coming from?"

Which Channels Actually Drive Revenue blends the numbers, flags the double-count, and recommends the safest reallocation with a confidence level.

"Here is our GTM export, the channel table, and the raw numbers for the board report. Give me the full picture before Monday."

Parallel Marketing Sweep runs the matched skills in parallel and hands back one ranked list, noting what it could not run for missing data.

You stay the decision-maker. The skills do the first pass.