Prompt Engineer Toolkit — Business

This skill turns marketing prompts into tested, versioned production assets. It provides A/B prompt evaluation against structured test cases, immutable prompt version history with diffs, and ready-to-use marketing prompt templates, plus an LLM-governance playbook for marketing teams.

marketing claude-code

📂 View skill on GitHub ↗ 🚀 How to use

✨ What it does

A/B prompt evaluation with quantitative scoring for adherence, relevance, and safety.
Prompt version tracking with immutable history, changelog, and diffs.
Reusable marketing prompt templates (ad copy, email, social, landing, SEO, brand voice).
Regression-friendly workflows for model/prompt updates.
LLM-governance playbook with claim discipline, disclosure rules, and human-review gates.
Evaluation rubric with mechanical scoring weights and acceptance gates.

🎯 When to use it

When launching a new LLM feature that needs reliable, measurable prompt outputs.
When prompt quality degrades after model or instruction changes and you need regression testing.
When multiple team members edit prompts and you need version history and diffs.
When you need evidence-based prompt selection for production rollout.
When you want consistent prompt governance across environments for AI-generated marketing content.

🚀 How to use

Trigger by mentioning 'prompt engineering', 'improve my prompts', 'prompt templates', 'prompt versioning', 'AI content workflow', or 'AI governance for marketing'. Provide your prompt drafts and test cases (JSON with input, expected_contains, forbidden_contains, expected_regex). Use the provided Python scripts to run A/B tests and version prompts. Example prompts:

Run an A/B test on my two ad copy prompts using the test cases in testcases.json.
Version my support_classifier prompt as v3 with author alice.

📄 Output: A/B test results, prompt version history and diffs, and access to marketing prompt templates and governance guidance.

📦 Add this skill to Claude Code

# 1. Get the skills repo
git clone --depth 1 https://github.com/alirezarezvani/claude-skills /tmp/claude-skills

# 2. Copy this skill into your project (or ~/.claude/skills for all projects)
mkdir -p .claude/skills
cp -r /tmp/claude-skills/marketing-skill/skills/prompt-engineer-toolkit .claude/skills/prompt-engineer-toolkit

Skill source: marketing-skill/skills/prompt-engineer-toolkit/SKILL.md

⚠️ Good to know

Requires Python 3 and an external LLM runner command for actual prompt evaluation; the skill provides scripts and templates but not the LLM itself.

❓ FAQ

How do I choose the winning prompt in an A/B test?

The tester scores outputs per case and aggregates metrics like content coverage, forbidden content violations, regex compliance, and length sanity. Choose the higher-scoring prompt as candidate baseline, then run a regression suite.

What should I avoid when editing prompts?

Avoid picking prompts from single-case outputs, changing prompt and model simultaneously, missing forbidden-content checks, editing without version metadata, skipping semantic diffs, optimizing one benchmark while harming edge cases, and model swaps without rerunning the baseline A/B suite.

What does the versioning policy require?

Use semantic prompt identifiers per feature, record author and change note for every revision, never overwrite historical versions, and diff before promoting a new prompt to production.

🤖 Overview, features, install steps and FAQ were generated from the project's SKILL.md on Sep 4, 2026. Always check the original source before running commands.