Lucio's Lovable Skills

A collection of reusable Lovable Skills built and used to make Lovable projects easier to review, understand, and improve — whether you are working alone, with a team, or with other AI tools.

What is a Skill?

A Skill is a saved set of instructions that Lovable can follow on command. Think of it like teaching Lovable a new trick: instead of typing the same long prompt every time, you just type the skill name and it knows exactly what to do.

Skills are different from Project Knowledge. Project Knowledge stores facts your app should remember. Skills store processes your app should run consistently.

Available skills

All skills are available at: lucioamor/lovable-skills

How to add a skill to Lovable

Open Lovable and go to:

Settings → Skills → Add → Import from GitHub

Paste the link for the skill you want:

https://github.com/lucioamor/lovable-skill-wireframe
https://github.com/lucioamor/lovable-skill-debate
https://github.com/lucioamor/lovable-skill-unbot
https://github.com/lucioamor/lovable-skill-jevify
https://github.com/lucioamor/lovable-skill-getfound

Lovable imports one skill at a time. You can add each skill separately.

Do not paste this catalog page URL. Use the individual skill links above.

Recommended workflow

Run /wireframe first to get a clear picture of what your app currently looks like.

Then run /debate when the next decision is not obvious, or when you want to find blind spots before committing. Then run /unbot when the chosen direction needs final copy that reads like a real person wrote it.

Run /getfound when the site needs to be found, understood and cited by search engines and AI beyond Lovable's native SEO review, and /jevify when the app makes runtime AI calls that might be cheaper or more reliable as structured decisions.

Example:

/wireframe

Then:

/debate based on WIREFRAME.md, why does the sign-up flow feel confusing?

Then:

/unbot [paste the chosen copy direction here]

This works especially well when switching between Lovable and external AI tools like ChatGPT, Claude, Gemini, or Grok — or when sharing context with a reviewer, client, or teammate.

Skill details

/wireframe

Maps the active page (or home page if unknown):

/wireframe

Maps one specific page:

/wireframe dashboard

Maps every page in the app into a single file:

/wireframe all

/wireframe reads the active page and writes a WIREFRAME-{page}.md file with every word of visible text, every button, every link, and every interactive element like filters and dropdowns. Use /wireframe all to map the full app into a single WIREFRAME.md.

When a page has very large tables or long blocks of text, the skill creates an abbreviated version and tells you the best format for reviewing the full data.

It does not design new screens, write placeholder text, judge the design, or make any changes to the app.

Full instructions: skills/wireframe/SKILL.md
Skill README: skills/wireframe/README.md
Example output: examples/WIREFRAME-nxlv.md

/debate

/debate [topic or question]

/debate pressure-tests a decision before you build it. It convenes specialists who argue opposing sides across different domains:

• convergent (one right answer matters) for technical architecture, security, data, and performance
• divergent (creative tension is useful) for design, copy, branding, and positioning
• mixed for user experience, where vision may differ but clarity and user protection still matter

It does not implement anything. It ends with trade-offs, a decision matrix, and a recommendation.

Full instructions: skills/debate/SKILL.md
Skill README: skills/debate/README.md

/unbot

/unbot [paste your text here]

/unbot rewrites prose to read as genuinely human-authored. It applies nine editing levers — word choice, sentence rhythm, hedge surgery, structural flattening, specificity, voice, human transitions, punctuation correction, and assistant-voice stripping — matched to the register of the text (marketing, email, social, docs, long-form).

Supply a writing sample and the skill will match your voice instead of a generic human default. Works in English and Brazilian Portuguese.

It does not invent facts, help text pass a detector, or touch meaning — form changes, claims don't.

Full instructions: skills/unbot/SKILL.md
Skill README: skills/unbot/README.md

/jevify

Run /jevify to audit runtime AI calls in your Lovable project. It returns an inventory, candidates for JEV, risks, and next steps in chat. It does not change code or reduce build credits.

Run /jevify migrate <finding> to move one candidate to JEV. It explains the plan, changes code only after you approve it, and adds the JEV decision next to the existing AI call behind an off | shadow | on flag that starts in shadow. Optionally add the jevify MCP server (https://jevify.lovable.app/mcp) under Connectors; without it, the skill runs locally.

• Full instructions: skills/jevify/SKILL.md.
• Skill README: skills/jevify/README.md.
• Claude Code variant and report template: lucioamor/jevify.

/getfound

/getfound [URL or scope]

/getfound works on the gap Lovable's native SEO & AI search review leaves open: entity identity and JSON-LD graph (@id, sameAs, knowsAbout), a ledger of claims with sources, what verified crawlers and unverified AI agents actually receive, Markdown fidelity, an AI-crawler policy by purpose, bilingual parity (hreflang, same facts in EN and PT-BR), agent files such as llms.txt and .well-known/agent-skills/, and a reproducible protocol to check what AI engines say and cite.

By default it audits and then applies the fixes that are local, reversible and backed by evidence. Run /getfound audit for a report only, or /getfound compare to re-measure against an earlier baseline. It never publishes, edits external profiles, invents facts, or promises rankings or citations.

Full instructions: skills/getfound/SKILL.md
Skill README: skills/getfound/README.md

Repositories

Versioning

Each skill tracks its own version. When you run a skill, it tells you the installed version and checks whether an update is available in its source repository.

Version status will show as:

• current — you have the latest version
• update available — a newer version exists; re-import the skill to update
• unverified — the check could not reach the source

Current published versions:

If a skill shows an older version, re-import its repository in Lovable to update it.

Maintainer workflow: MAINTENANCE.md

Authorship and maintenance

This project was created by Lucio Amorim, Lovable Partner.

When reusing, redistributing, or citing this work, keep the attribution credits and include a link to this repository.

License

This catalog and the standalone skill repositories are licensed under
Creative Commons Attribution 4.0 International (CC BY 4.0).

That means you may copy, share, adapt, remix, publish, and use these skills,
including commercially, as long as you give appropriate credit to
Lucio Amorim, link to the license,
and indicate whether you made changes.

In plain terms: you can use the skills freely, but attribution is required.

How these skills are built (technical reference)

This section is for developers, contributors, or curious readers who want to understand how the skill system works under the hood.

What a Skill is, technically

A Lovable Skill is a Markdown file (SKILL.md) with a YAML frontmatter block at the top and a set of instructions written in plain English below it. Lovable reads the frontmatter to register the skill's name and a short description, and reads the instructions to know how to behave when the skill is triggered.

Example frontmatter:

---
name: wireframe
description: Run /wireframe to read your app and write WIREFRAME.md...
---

The instructions below the frontmatter are written so that an AI can follow them reliably: they define the trigger, the workflow steps, the output format, and any edge-case rules.

Repository structure

This catalog repository stores the authoritative source for each skill:

lovable-skills/
├── skills/
│   ├── wireframe/
│   │   ├── SKILL.md       ← authoritative skill instructions
│   │   ├── README.md      ← user-facing documentation
│   │   └── VERSION.md     ← current version
│   ├── debate/            ← same three files in every skill folder
│   ├── unbot/
│   ├── getfound/
│   └── jevify/            ← generated mirror of lucioamor/jevify (+ LICENSE); do not edit here
├── examples/              ← sample output files
├── LICENSE                ← CC BY 4.0, copied into every standalone repo
├── MAINTENANCE.md         ← maintainer workflow
└── SKILL_METADATA.md      ← description contract

Standalone import repositories

Lovable's import tool requires a GitHub repository with a SKILL.md at the root. Because of this, each skill also lives in its own standalone repository:

• lucioamor/lovable-skill-wireframe
• lucioamor/lovable-skill-debate
• lucioamor/lovable-skill-unbot
• lucioamor/lovable-skill-jevify
• lucioamor/lovable-skill-getfound

These standalone repositories are export targets, not editing targets. The canonical source lives in skills/<skill-name>/SKILL.md in this catalog. When a skill is updated here, the corresponding standalone repository should be synced.

The export contract is documented in MAINTENANCE.md.

Versioning convention

Each skill follows simple semantic versioning (v1.0.0). The version is stored in VERSION.md inside the skill folder and referenced inside SKILL.md. When a skill runs, it compares its installed version against the standalone repository's VERSION.md to report whether an update is available.

How to add a new skill

1. Create skills/<new-skill>/SKILL.md with valid frontmatter and instructions.
2. Create skills/<new-skill>/VERSION.md starting at v1.0.0.
3. Create skills/<new-skill>/README.md for user-facing documentation.
4. Update README.md (this file) and SKILL_METADATA.md.
5. Create a new standalone import repository and sync the files there.
6. Add the new repository link to the Repositories table above.