ai-copywriter: The Prompt That Treats Marketing Copy Like a Build System

A portable skill that drafts, audits, and humanizes prose with intake gates, version checks, and a 33-pattern cleanup loop.

8 min read • View on GitHub • More from mikiarlo3

A split editorial desk scene shows one side flooded with generic pages spilling from a machine, while the other side passes the pages through a human editor’s checklist and red pencil into a clean final sheet. The image explains that the repo is not just generating copy, it is controlling the whole writing process.
The prompt is less a writer than a controlled editorial pipeline.
Key Takeaways

The editor is the product

Most AI copy tools try to be faster than a writer. This repo tries to be stricter than an editor. That shift matters, because the core promise is not volume, it is control: intake first, draft second, audit third, final only if the piece survives the cleanup loop.

It is plain Markdown, so it runs in any harness that supports skill-style instructions.

Project README, Repository documentation · mikiarlo3/ai-copywriter README

That framing turns the skill into a writing system. The model is not asked to improvise its way to good copy. It is asked to behave like a copywriter, then like a ruthless reviewer, then like a final pass that only ships the parts that still sound human.

Why it refuses to write too early

The intake step is doing real work. It pressures the user for story, audience, offer, and context before generating prose, which prevents the usual failure mode where a vague prompt produces polished nonsense.

The skill is built to refuse weak input, then route strong input through a correction loop.

That is the difference between a prompt and a process. A prompt can ask for copy. A process can decide whether the brief is worthy of copy, and if not, stop there.

Draft, audit, final

This is the technical center of gravity. The first pass generates marketing structure. The second pass scans for machine tells: over-structured phrasing, inflated symbolism, awkward parallelism, and the little habits that make copy feel synthetic.

StageGeneric AI writerai-copywriter
InputAccepts vague prompts and fills the gapsPressures the brief until the story, audience, and offer are clear
DraftingOptimizes for speed and completionUses explicit marketing structure
RevisionOften one more generic rewriteRuns an audit that targets human-ness
HumanizationUsually a style promptA checklist of specific writing faults
Failure modeConfident slopRefusal until the brief is usable
A close-up manuscript page passes under a magnifying glass that reveals a matrix of writing faults being checked off. Small editorial marks suggest specific issues are being removed as the page moves through the lens. The image explains how <span class=the audit phase turns subjective cleanup into a repeatable filter." data-prompt="Create an editorial illustration rendered entirely in black ink on a pure white background. The image depicts a close-up manuscript page being scanned by a magnifying glass. The magnifying glass reveals a stack or grid of small editorial checks for writing faults, with tiny visual cues suggesting over-structured phrasing, inflated symbolism, and repetitive cadence being removed as the page moves under the lens. The manuscript page is actively cleaned as it passes beneath the glass, with a few lines crossing out and replacing awkward phrases. The artist uses tightly packed crosshatching lines layered at different angles to build up shadow and form, with clean open areas of white for highlights. The line work has the quality of a classic metal engraving, precise and deliberate, with varied line weights where bold contour lines define shapes and finer interior lines create tonal depth. The overall style evokes vintage newspaper editorial illustrations from The Economist or Wall Street Journal. No color, no gradients, no grey fills. Only black lines on white. The background MUST be pure white #FFFFFF. No paper texture, no cream, no off-white, no noise, no grain. Perfectly clean flat white background. The composition is a close-up. The magnifying glass dominates the frame and the manuscript fills the rest of the scene." loading="lazy">
The audit phase is not a vague style polish. It is a targeted scan for writing habits that feel machine-made.

The project’s moat is not that it can write. Plenty of models can write. The moat is that it can diagnose bad writing in a structured way, then make the next pass materially better.

The 33-pattern cleanup is the real moat

The README says the skill is built on Humanizer, which packaged Wikipedia’s Signs of AI writing into 33 detectable, fixable patterns. That matters because it converts a subjective problem into something operational.

It is built on blader's Humanizer, which packaged Wikipedia's "Signs of AI writing" guide into 33 detectable, fixable patterns.

Project README, Repository documentation · mikiarlo3/ai-copywriter README

Instead of asking the model to become “more human,” the repo gives it a taxonomy of failure. That is a much better engineering move. It lets the system correct for specific habits instead of chasing a vibe.

Why portability matters more than novelty

This is not a one-off prompt dump. The repo is packaged across harnesses, with version sync and validation scripts that keep the manifests aligned. That makes it feel like software, not a clever paste into a chat window.

# scripts/validate-package.py
# Ensures version strings match across SKILL.md, README.md, and plugin.json
# Fails if the prompt package gets too large for portability

if skill_version != readme_version or readme_version != plugin_version:
    raise SystemExit('Version mismatch')

if skill_lines > 600:
    raise SystemExit('Prompt exceeds portability budget')
ConcernGeneric writing toolai-copywriter
PackagingLives inside one productShips as portable skill-style instructions
VersioningUsually opaqueExplicit sync across files
ValidationAd hocScripted checks before release
Context limitsIgnored until it breaksManaged as a design constraint
ReuseHard to lift into another harnessBuilt for cross-harness portability

That packaging choice changes the category. The repo is not selling inspiration. It is selling repeatable editorial infrastructure.

What this changes about AI copywriting

The broader lesson is simple: the best AI writing tools may stop trying to be authors and start acting like editorial systems. If you care about conversion, tone, and trust, the interesting question is no longer whether the model can draft. It is whether the workflow can keep the model honest.

QuestionGeneric AI copywritingai-copywriter
What is it optimizing?Speed and outputReader empathy and tone control
Who is in charge?The modelThe editorial workflow
What fixes the text?One more rewriteA structured audit loop
What makes it useful?ConvenienceProcess discipline