en un clic
social-copy-guardrails
// Enforce Antonio's real Spanish social voice when drafting or rewriting X, LinkedIn, article teaser, and newsletter intro copy. Use whenever copy must sound like Antonio and avoid obvious AI writing.
// Enforce Antonio's real Spanish social voice when drafting or rewriting X, LinkedIn, article teaser, and newsletter intro copy. Use whenever copy must sound like Antonio and avoid obvious AI writing.
Process invoice PDFs/emails: extract vendor + date, generate normalized filename, optionally archive to Drive, and email the PDF to the right Holded inbox.
Use the private LearnWorlds CLI to inspect DevExpert Academy users, find students by email, list their enrolled courses/products, look up products, and perform safe enrollment workflows. Trigger when Antonio asks what courses a student has in LearnWorlds or academia.devexpert.io, whether someone belongs to the current or next AI Expert edition, or to use the LearnWorlds/academy CLI.
Genera la newsletter semanal de DevExpert recopilando contenido de X, YouTube, PostFlow y bookmarks. Crea borrador en Listmonk para revisión.
End-to-end YouTube publishing workflow using ordered scripts: prepare/concat video, upload draft, transcribe with Parakeet, generate copy with the calling model, optionally prepare English dubbing assets, render thumbnails, update YouTube metadata, then schedule socials (PostFlow) 15 minutes after publish.
Converts a Markdown file to a styled PDF with DevExpert branding (logo in bottom-right corner). Use when asked to generate a PDF from a Markdown document, or when any DevExpert proposal/document needs to be exported as PDF.
Manage inbox email. Uses the inbox script and stores metadata (ids) to open or archive messages later.
| name | social-copy-guardrails |
| description | Enforce Antonio's real Spanish social voice when drafting or rewriting X, LinkedIn, article teaser, and newsletter intro copy. Use whenever copy must sound like Antonio and avoid obvious AI writing. |
This is a living skill for writing social copy in Antonio's voice.
If Antonio corrects any text generated with this skill, update this skill or one of its resources so the correction becomes part of the workflow from that point on. Do not treat corrections as one-off preferences for the current turn.
Use this skill whenever the task involves:
This skill is not the default tool for content strategy. It is for turning a real idea, experience, draft, or scene into copy that sounds like Antonio.
Optimize for voice first.
The default goal is not maximum authority, maximum reach, or a perfect creator post. The default goal is: Antonio should be able to publish the result after changing at most two or three things.
Good copy here should feel like:
If a draft sounds strategically correct but not like Antonio, it is wrong.
Before drafting or rewriting copy, read:
resources/copy-workflow.mdresources/voice-profile.mdresources/recent-manual-x-corpus.mdresources/feedback-bank.mdUse resources/ai-writing-tropes.md only as a final lint reference when the
draft feels too polished, too symmetrical, or too obviously generated. Do not
load it as primary generative context.
Do not read ~/Documents/aipal/10-areas/contenido/patrones-redes.md by default.
Only use that strategy file when Antonio explicitly asks for distribution,
performance, calendar, positioning, or content-system decisions.
Follow the workflow in resources/copy-workflow.md.
The short version:
If Antonio says something like "escríbelo ya", "dame el texto", "programalo", "publícalo", or provides a nearly final draft, skip the angle step and produce the final text directly.
Do not erase Antonio's wording just to make the copy smoother.
Before drafting, identify the phrases, hesitations, observations, or exact turns that should survive. If the input contains a human line like "no sé cómo contarlo", "me dejó impresionado", "me cuesta bastante poco detectar que es suyo", or "pues eso", preserve that kind of texture unless it clearly hurts the post.
When the raw material is a personal experiment, workflow session, or anecdote, do not force it into a lesson about authority, positioning, or what it proves for the whole industry.
Prefer:
Avoid endings like:
esto demuestra la profundidad que se puede alcanzaresto cambia el ciclo de feedbackpoder delegar elimina la friccionasi es como deberiamos trabajar a partir de ahoraAntonio's recent manual X posts often work because they feel like thinking in public. Keep some natural asymmetry when it belongs to the idea.
Avoid making every post:
Some posts can start with uncertainty, a scene, or a conversational admission.
Do not turn every post into positioning.
Words like "autoridad", "impacto", "referente", "comunidad", "tesis", and "posicionamiento" belong to strategy conversations, not to most publishable copy. If those ideas are not in Antonio's input, do not introduce them.
Avoid stacking common AI tells:
no es X, sino Ylo importante no es X, es Yla pregunta útil no es Xesto demuestra...la clave está en...en un mundo donde...lo que nadie está viendo...no se trata solo de...One contrast can work when Antonio used it or when it genuinely sharpens the point. Repeated contrast framing makes the copy sound generated.
Do not imply Antonio assumed a tool was a toy, fake demo, headline, or gimmick unless he explicitly framed it that way.
Avoid recurring assistant-default lines like:
no es una demo de jugueteno hace falta convertirlo todo en una gran demono solo mirarlo desde el titulardónde empieza a molestarla pelea interesante está en...me interesaba especialmente ver cómo se comportaba...Write like a Spanish-speaking software developer who lives inside these tools. Not academic, not corporate, not creator-economy smooth.
Allowed when natural:
puesa veren lo personalseguramenteyo qué sées curiosode locosme cuesta bastanteDo not sprinkle these mechanically. Use them only where they fit.
Antonio often uses numbered emoji lists naturally, especially for practical or opinion posts.
When a list improves the copy:
1️⃣, 2️⃣, 3️⃣ are fine for ordered points👉 is fine for simple examples✅, ❌, and ⚠️ are fine when they match the meaning•Do not force a list if the idea works better as prose.
When the topic is a library, framework, model, benchmark, launch, or repo:
When the post is meant to send readers to a video:
For subscriber-facing copy, avoid mentioning internal publishing or ops tools unless they are part of the actual story for the audience.
Use reader-facing wording instead of internal implementation detail.
When Antonio asks for copy:
When Antonio provides his own draft:
If Antonio corrects tone, structure, recurring phrases, list formatting, cadence, or technical nuance:
resources/feedback-bank.mdresources/voice-profile.md if it changes the durable voice modelresources/recent-manual-x-corpus.md
only when it is useful as future training contextA good output: