convez.ai
SkillIntegração & APIsIntermediário
PhAlves23

Prompt Cause Marketplace

PhAlves23/promptcause-marketplace

Marketplace curado de skills de IA para Claude Code e agentes compatíveis

Ver no GitHub
BomPluginScore · de 0 a 100

Um marketplace curado de AI skills para Claude Code e outros agentes de IA. Permite instalar capabilities focadas (skills) a partir de um catálogo centralizado, com revisão e versionamento seguro. Uma forma organizada de compartilhar e reutilizar prompts e workflows.

Instalação

$npx convez-ai PhAlves23/promptcause-marketplace

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
3
Forks
1
Acompanhando
0
Saúde
100%
Último commit
há 3 meses
Versão
—
Issues abertas
1
Contribuidores
2

Casos de uso

  • Instalar skills de engenharia de prompts para melhorar workflows
  • Gerenciar múltiplas capabilities de IA a partir de um único catálogo
  • Compartilhar e descobrir skills verificadas para Claude Code

Veredito

Projeto bem estruturado que resolve um problema real: centralizar e versionär skills de IA de forma segura. A proposta é sólida (marketplace vs repo individual), mas ainda está em estágio inicial com apenas 1 skill no catálogo. Requer ecossistema robusto de contribuidores para virar realmente útil.

README

A tradução para o português deste README ainda está na fila.

Prompt Cause — Skill Marketplace

A curated catalog of AI skills for Claude Code and other agents. Add one marketplace, install any skill.

License: MIT Validate Skills Security: sha-pinned PRs Welcome

Works with Claude Code · GitHub Copilot · OpenAI Codex · Gemini CLI · OpenCode · Cursor · Windsurf · Cline


Table of contents


What is this

An AI skill is a focused capability you can drop into an AI coding agent — a SKILL.md plus reference files that teach the agent to do one thing well, and to know when to do it.

Prompt Cause is a curated marketplace of those skills. Instead of hunting down individual repos and copying files around, you add one marketplace and install any skill from a single catalog — with a review-and-pin process that keeps third-party skills safe to run.

Why a marketplace instead of one repo per skill?

  • One source for users. Add it once; the whole catalog is available.
  • Granular. Install only the skills you want — nothing is bundled or forced.
  • Open but curated. Anyone can submit a skill; every listing is reviewed and version-pinned before it ships.

Quick start

Claude Code

/plugin marketplace add PhAlves23/promptcause-marketplace
/plugin install <skill-name>

Restart Claude Code, then use the skill (e.g. say "improve this prompt" for prompt-engineering).

GitHub Copilot CLI

copilot plugin marketplace add PhAlves23/promptcause-marketplace
copilot plugin install <skill-name>@promptcause

For Cursor, Windsurf, Codex, Gemini, and OpenCode, each skill's own repo ships native install instructions — see the catalog and Using skills in other tools.

Catalog

SkillWhat it doesCategoryRepo
prompt-engineeringRewrites a raw draft into a production-grade prompt, with a changelog of what changed and why. Grounded in primary sources (Anthropic, OpenAI, Google, The Prompt Report).productivityprompt-engineering-skill

Want your skill here? See Contributing.

How it works

Claude Code (and compatible agents) use a three-level hierarchy:

Marketplace  →  contains many Plugins  →  each Plugin contains one or more Skills
   (this repo)        (a catalog entry)          (the actual SKILL.md)

This repository is the marketplace: a catalog defined in .claude-plugin/marketplace.json that points at skills. A skill can live in one of two places:

Locationsource in the catalogBest for
Its own repo (recommended)git-subdir / github / url, pinned to a shaSkills maintained independently, incl. community ones
Inline in this repo"./plugins/<name>"Small skills donated to the catalog

Either way, users add a single marketplace and install skills individually.

Security model

Listed skills run inside other people's agents, so safety is built into the process — not an afterthought. Three layers:

  1. Review before listing. Every submission is audited for hidden instructions (prompt injection, data exfiltration, covert tool/URL calls), unsafe scripts, and scope that doesn't match its description.
  2. Pinned commits. Externally-hosted skills are pinned to a reviewed commit sha. A contributor can change their own repo freely, but those changes do not reach users here until a new sha is reviewed and merged.
  3. Re-review on update + CI enforcement. Updating a skill is a new PR changing the sha (re-audited). The validate workflow rejects any external source that isn't pinned to a full commit sha.

The repo itself is hardened too: least-privilege CI permissions, GitHub Actions pinned by sha, CODEOWNERS review, and Dependabot. Full policy in SECURITY.md and CONTRIBUTING — Versioning & security.

Using skills in other tools

Skills here target Claude Code first, but the format is portable. Each skill's repo documents its own native install for:

ToolMechanism
Claude Code / Copilot CLIThis marketplace (/plugin or copilot plugin)
OpenAI CodexNative skill discovery (~/.agents/skills)
Gemini CLIExtension
Cursor / Windsurf / ClineRules files (.cursor/rules, .windsurf/rules, .clinerules)
OpenCodeAGENTS.md / skills path

Contributing a skill

Community skills are welcome. Two paths:

  1. Keep your skill in your own repo (recommended) — you own and maintain it; we list it (pinned to a reviewed sha).
  2. Contribute it inline — add it under plugins/ in this repo.

Start from skill-template/, then follow CONTRIBUTING.md. Every submission is reviewed for quality and security before listing.

FAQ

Every listing is reviewed and pinned to a specific reviewed commit, so post-review changes in a contributor's repo can't silently reach you. That said, a skill is third-party content — install skills you trust, and review the source repo if in doubt. No marketplace can guarantee third-party code is safe; this is defense-in-depth. See SECURITY.md.

/plugin update <skill-name> (Claude Code) or copilot plugin update. You get the latest reviewed version — i.e. whatever sha is currently merged in this catalog.

/plugin uninstall <skill-name> removes a skill. /plugin marketplace remove promptcause removes the catalog.

Yes — see Using skills in other tools. Each skill's repo has native instructions for Cursor, Windsurf, Codex, Gemini, OpenCode, and Cline.

No — and that's by design. Your listing is pinned to the sha that was reviewed. To ship an update, open a PR here changing the sha; it's re-reviewed before merge. This protects users from unreviewed changes. See CONTRIBUTING — Versioning & security.

Skills are designed to be lean: a small always-on footprint (the description + triggers), with heavier reference material loaded only when the skill actually fires. Check a skill's details (claude plugin details <name>) for its token cost.

Repository structure

promptcause-marketplace/
├── .claude-plugin/marketplace.json   # the catalog (skills + pinned sources)
├── plugins/                          # inline skills (Path B), if any
├── skill-template/                   # copy this to start a new skill
├── scripts/bump_skill.py             # advance owned skills to their latest sha
├── assets/                           # banner (png + editable html source)
├── docs/ARCHITECTURE.md              # how the catalog, pinning, CI and bump work
├── .github/
│   ├── workflows/validate.yml        # JSON + catalog + sha-pin enforcement
│   ├── workflows/bump-own-skills.yml # opens a PR bumping owned skills
│   ├── ISSUE_TEMPLATE/submit_skill.yml
│   └── CODEOWNERS · dependabot.yml · PULL_REQUEST_TEMPLATE.md
├── CONTRIBUTING.md · SECURITY.md · CODE_OF_CONDUCT.md · CHANGELOG.md · LICENSE

Roadmap

  • Grow the catalog with more first-party and community skills
  • Per-skill category browsing as the catalog grows
  • Automated source-resolves check in CI (verify each pinned sha exists upstream)
  • A simple website listing the catalog

Have an idea? Open a discussion.

License

MIT for the catalog and tooling in this repo. Each externally-hosted skill keeps its own license.

Ver o original no GitHub

Bom para

Relacionados