Claude Skill

nemawashi

Clarify software goals, constraints, stakeholders, and risks before consequential design or implementation work. Use for greenfield ideas, ambiguous requirements, architecture decisions, or changes where misunderstanding would be costly.

LLM Mart · 0 points · 0 views 0 listing impressions 0 install-command copies
Virus-scanned Reviewed automatically before listing.

Full trust report

Download kuya-egg-monozukuri-nemawashi-b4243fe.zip · 2 KB
Part of kuya-egg/monozukuri — 13 skills

Install

skills CLI npx skills add https://github.com/kuya-egg/Monozukuri/tree/main/nemawashi
Claude Code claude plugin marketplace add https://llmmart.ai/marketplace.json && claude plugin install kuya-egg-monozukuri@llmmart
Git git clone https://github.com/kuya-egg/Monozukuri.git

The skills CLI installs just this skill, for any of its supported agents. Claude Code installs the whole kuya-egg/monozukuri collection as a plugin from our marketplace. Git is the plain clone.

Skill manifest

Nemawashi

Nemawashi is the practice of preparing the ground before a consequential change. In software, it means building shared understanding before building the system. This is a Japanese-inspired engineering metaphor, not a claim about how every team works.

Use this skill when

  • the request is ambiguous or has competing interpretations;
  • a new project is starting from scratch;
  • architecture, data shape, public API, migration, or operational behavior could be difficult to reverse;
  • multiple people, systems, or teams depend on the outcome; or
  • implementation would otherwise require guessing.

Do not use a long interview for a clear, low-risk edit. Ask only questions whose answers can change the work.

Outcome

Produce a concise engineering brief containing:

  • the problem and why it matters;
  • users, systems, or stakeholders affected;
  • desired behavior and explicit non-goals;
  • constraints and success criteria;
  • facts, decisions, assumptions, and open risks separated clearly;
  • the smallest useful next step; and
  • how the result will be verified.

Workflow

1. Establish the frontier

State what is already known and identify the first uncertainty that could change the design. Do not ask for information that will not affect the next decision.

2. Interview deliberately

Ask one focused question at a time. Prefer questions about:

  • the user-visible outcome;
  • the boundary of the request;
  • data ownership and lifecycle;
  • compatibility and migration needs;
  • failure tolerance and recovery;
  • security and authorization;
  • performance or scale that is actually required; and
  • what “done” means.

When a reasonable default exists, state it and continue rather than creating an endless questionnaire. Ask for confirmation when the default changes risk, scope, or external behavior.

3. Make the shape explicit

Summarize the problem in plain language. Name what will change, what will not change, and which decisions are reversible. For a greenfield project, include the first vertical slice rather than designing the entire future platform.

4. Stop at the right boundary

Stop interviewing when the next work can be planned and verified without guessing. If a critical question remains unresolved, stop before implementation and explain why it matters.

Evidence standard

Do not present an assumption as a fact. Label statements as Observed, Decided, Assumed, or Open. For an existing project, point to the files, commands, documentation, or behavior that support important claims.

Boundaries

  • Do not make architecture decisions merely to end the interview.
  • Do not expand a feature into a platform because future use is imaginable.
  • Do not substitute consensus theater for a concrete decision.
  • Do not block a reversible prototype on questions that only matter at scale.
  • Do not implement while a material ambiguity remains hidden.

Handoff

End with a short brief another engineer can act on. Recommend the next focused skill only when useful: genchi-genbutsu for evidence, kanso for design, or kata for an already settled implementation.

If a referenced skill is not installed, apply its named lens inline instead of trying to invoke it.

Files (monozukuri)
  • agents
    • openai.yaml 340 B
      interface:
        display_name: "Nemawashi — Clarify Before Changing"
        short_description: "Clarify requirements before consequential changes"
        brand_color: "#1F2937"
        default_prompt: "Use $nemawashi to clarify this software idea, its constraints, and its success criteria before implementation."
      
      policy:
        allow_implicit_invocation: false
      
  • SKILL.md 3.4 KB
    ---
    name: nemawashi
    description: "Clarify software goals, constraints, stakeholders, and risks before consequential design or implementation work. Use for greenfield ideas, ambiguous requirements, architecture decisions, or changes where misunderstanding would be costly."
    ---
    
    # Nemawashi
    
    Nemawashi is the practice of preparing the ground before a consequential
    change. In software, it means building shared understanding before building
    the system. This is a Japanese-inspired engineering metaphor, not a claim
    about how every team works.
    
    ## Use this skill when
    
    - the request is ambiguous or has competing interpretations;
    - a new project is starting from scratch;
    - architecture, data shape, public API, migration, or operational behavior
      could be difficult to reverse;
    - multiple people, systems, or teams depend on the outcome; or
    - implementation would otherwise require guessing.
    
    Do not use a long interview for a clear, low-risk edit. Ask only questions
    whose answers can change the work.
    
    ## Outcome
    
    Produce a concise engineering brief containing:
    
    - the problem and why it matters;
    - users, systems, or stakeholders affected;
    - desired behavior and explicit non-goals;
    - constraints and success criteria;
    - facts, decisions, assumptions, and open risks separated clearly;
    - the smallest useful next step; and
    - how the result will be verified.
    
    ## Workflow
    
    ### 1. Establish the frontier
    
    State what is already known and identify the first uncertainty that could
    change the design. Do not ask for information that will not affect the next
    decision.
    
    ### 2. Interview deliberately
    
    Ask one focused question at a time. Prefer questions about:
    
    - the user-visible outcome;
    - the boundary of the request;
    - data ownership and lifecycle;
    - compatibility and migration needs;
    - failure tolerance and recovery;
    - security and authorization;
    - performance or scale that is actually required; and
    - what “done” means.
    
    When a reasonable default exists, state it and continue rather than creating an
    endless questionnaire. Ask for confirmation when the default changes risk,
    scope, or external behavior.
    
    ### 3. Make the shape explicit
    
    Summarize the problem in plain language. Name what will change, what will not
    change, and which decisions are reversible. For a greenfield project, include
    the first vertical slice rather than designing the entire future platform.
    
    ### 4. Stop at the right boundary
    
    Stop interviewing when the next work can be planned and verified without
    guessing. If a critical question remains unresolved, stop before
    implementation and explain why it matters.
    
    ## Evidence standard
    
    Do not present an assumption as a fact. Label statements as `Observed`,
    `Decided`, `Assumed`, or `Open`. For an existing project, point to the files,
    commands, documentation, or behavior that support important claims.
    
    ## Boundaries
    
    - Do not make architecture decisions merely to end the interview.
    - Do not expand a feature into a platform because future use is imaginable.
    - Do not substitute consensus theater for a concrete decision.
    - Do not block a reversible prototype on questions that only matter at scale.
    - Do not implement while a material ambiguity remains hidden.
    
    ## Handoff
    
    End with a short brief another engineer can act on. Recommend the next
    focused skill only when useful: `genchi-genbutsu` for evidence, `kanso` for
    design, or `kata` for an already settled implementation.
    
    If a referenced skill is not installed, apply its named lens inline instead of
    trying to invoke it.
    

Comments (0)

Sign in to join the conversation.

No comments yet.

Reviews (0)

No reviews yet.

Related