convez.ai
SkillProdutividadeIntermediário
adelaidasofia

Claude Deconstruct

adelaidasofia/claude-deconstruct

Análise de primeiros princípios que revela o que a análise convencional não vê

Ver no GitHub
InicialPluginScore · de 0 a 100

Skill de análise de primeiros princípios que expõe pressupostos ocultos, estabelece verdades fundamentais e identifica o movimento de maior alavancagem. Segue o método aristotélico para reconstruir problemas do zero, descartando convenção e estratégia anterior.

Instalação

$npx convez-ai adelaidasofia/claude-deconstruct

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
1
Forks
0
Acompanhando
0
Saúde
42%
Último commit
há 1 dia
Versão
—
Issues abertas
0
Contribuidores
2

Casos de uso

  • Questionar modelos de precificação copiados de competidores
  • Validar roadmaps de produto contra necessidades reais vs. desejos declarados
  • Avaliar decisões estratégicas de carreira e negócio sem viés de convenção
  • Identificar pressupostos carregados de medo em decisões de alto risco

Veredito

Skill robusto e bem pensado para quem enfrenta decisões complexas. O método em quatro fases é sólido; a detecção de viés de medo é diferencial honesto. Recomendado para líderes, estrategistas e founders que querem raciocínio estruturado além do convencional.

README

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

claude-deconstruct

A Claude Code plugin for first-principles analysis. Surfaces hidden assumptions, establishes foundational truths, rebuilds the problem from scratch, and identifies the high-leverage move that conventional analysis would miss.

Modeled on Aristotle's method: find the truths that cannot be derived from anything more basic, then reason upward from those alone.


What it does

You run /deconstruct and describe a problem, decision, or situation. The skill walks four phases:

  1. Surface the assumptions — what's embedded in how you framed it (and which are load-bearing)
  2. Establish first principles — what's true independent of convention or prior strategy
  3. Rebuild from the foundation — three distinct approaches, optimized for speed, impact, simplicity
  4. The high-leverage move — the single action that's invisible under conventional analysis

Two modes:

  • Fast (Phase 1 + Phase 4) — for daily decisions, auto-triggers
  • Full (all four phases) — for pricing, business models, career decisions, strategy pivots

The fear flag is built in: when an assumption traces back to fear, the skill names it and points to journaling instead of strategy. Fear-origin assumptions are rarely intellectual problems.


Install

Open Claude Code, paste:

/plugin marketplace add adelaidasofia/claude-deconstruct
/plugin install claude-deconstruct@claude-deconstruct

Then in any Claude Code session:

/deconstruct

The skill will prompt for the problem you want to deconstruct.

claude plugin add github.com/adelaidasofia/claude-deconstruct

Requirements

  • Claude Code
  • Nothing else

Optional integrations:

  • Decision template with stakes: high field — auto-offers fast mode on high-stakes entries (set deconstruct_auto: true in the decision frontmatter to enable)
  • The daily-journal or insights plugin — adds a "first-principles audit" step to weekly review (flags high-stakes decisions made without a deconstruct pass)

When to run it

This skill works best when you suspect you're stuck inside assumptions you can't see:

  • A pricing model copied from competitors without questioning why
  • A product roadmap built on what users say they want vs. what they actually need
  • A career path followed because "that's how it's done in this industry"
  • A business model that feels stuck but can't figure out why
  • A hiring process, marketing strategy, or workflow everyone told you was "best practice"
  • A competitive analysis where you're parroting positioning instead of finding your edge
  • Any decision where "we've always done it this way" is the real answer

The full run is heavy. Don't run it on trivial decisions. Fast mode exists for a reason.


What makes this different from "list pros and cons"

Pro/con lists assume the framing is correct. Deconstruction questions the framing.

Most decision tools take your problem statement as given and help you optimize within it. This skill assumes the problem statement itself is contaminated by convention, imitation, precedent, fear, and unexamined defaults — and forces you to strip those out before you reason about solutions.

The 3-question test in Phase 2 is the core. If a statement is only true because everyone else does it, it's not a first principle. It's a habit.


Submitting to the Cowork marketplace

Once the repo is on GitHub, submit at: clau.de/plugin-directory-submission

For the community marketplace:

claude plugin marketplace add anthropics/claude-plugins-community

Telemetry

This plugin sends a single anonymous install signal to myceliumai.co the first time it loads in a Claude Code session on a given machine.

What is sent:

  • Plugin name (e.g. slack-mcp)
  • Plugin version (e.g. 0.1.0)

What is NOT sent:

  • No user identifiers, names, emails, tokens, or API keys
  • No file paths, message content, or anything from your work
  • No IP address is stored after dedup processing

Why: Helps the maintainer know which plugins people actually install, so attention goes to the ones that get used.

Opt out: Set the environment variable MYCELIUM_NO_PING=1 before launching Claude Code. The hook will skip the network call entirely. Already-pinged installs leave a sentinel at ~/.mycelium/onboarded-<plugin> — delete it if you want to reset state.

License

MIT

Ver o original no GitHub

Bom para

Relacionados