Software Concept Architect · Refine for Openclaw

An evidence-led coding-agent skill that finds the smallest justified change to improve a software concept or concept system.

agenticweb4
v1.0.0
Sep 30, 2026
0
145
0

Install & Download

1. ClawHub CLI

The fastest way to install a skill directly from the registry.

npx clawhub@latest install software-concept-architect-refine

2. Manual Installation

Copy the skill folder to one of these locations

Global
~/.openclaw/skills/
Workspace
<project>/skills/

Priority: Workspace > Local > Bundled

3. Prompt Installation

Copy this prompt to OpenClaw to install it automatically.

Help me install software-concept-architect-refine using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).

Prefer to download?

Get the raw skill files in a ZIP archive.

What is Software Concept Architect · Refine?

Software Concept Architect · Refine helps an AI coding agent analyze existing concepts against their intended purpose, using requirements, specifications, relevant code, real scenarios, and counterexamples. It investigates problems such as overload, redundancy, excessive narrowness, fragmentation, and synchronization that is either too loose or too strict.

Rather than defaulting to a redesign, the skill compares possible moves—including splitting or merging concepts, unifying or specializing them, and tightening or loosening synchronization—against the root cause. It explains trade-offs and impact, distinguishes facts from inferences, and supports Openclaw Skills users in making deliberate, verifiable changes while preserving behavior that should remain unchanged.

Software Concept Architect · Refine Use Cases

  • A concept or concept system has accumulated too many responsibilities and may be overloaded.
  • Multiple concepts appear redundant, or a concept may be too narrow to fulfill its purpose.
  • A design has fragmented across components, making responsibilities or behavior difficult to understand.
  • Synchronization between concepts or components seems excessive, insufficient, or inconsistent with their contracts.
  • A proposed split, merge, unification, or specialization needs to be evaluated against actual evidence.
  • Changes have been agreed upon and need implementation with preserved behavior, intentional differences, and verification clearly distinguished.

How Software Concept Architect · Refine Works

  1. Read the requirements, specifications, and relevant code to establish the concept system’s context.
  2. Reconstruct each relevant concept’s purpose, state, actions, operational principle, and synchronization relationships.
  3. Investigate suspected problems using actual scenarios and counterexamples; gather facts and label inferences.
  4. Read the required moves and evidence reference before choosing a design move.
  5. Compare candidate responses—split or merge, unify or specialize, tighten or loosen synchronization—and include no change as a valid option.
  6. Explain the expected benefits, costs, and scope of impact. Discuss consequential choices in rounds, after their prerequisites are settled, and provide recommendations with reasons.
  7. Once the approach is settled and authorized, make the agreed changes, distinguishing preserved behavior from intentional changes.
  8. Verify local purposes and end-to-end scenarios, stop when the agreed goal is met without known major regressions, and identify anything that remains unverified.

Software Concept Architect · Refine Setup

The supplied SKILL.md does not specify a package manager, installation command, or agent-specific installation path. Add the skill to the skills location supported by your coding agent, preserving its SKILL.md and the referenced references/moves.md file. Make the relevant requirements, specifications, and code available to the agent before requesting an analysis.

Invoke the skill when you want an existing software concept or concept system evaluated and refined. Agree on the proposed approach before authorizing implementation; the skill calls for consequential choices to be discussed before changes are made.

Software Concept Architect · Refine Data Schema & Taxonomy

The skill defines guidance and metadata rather than an application data model or generated output files.

Item Organization and purpose
SKILL.md frontmatter YAML metadata identifies the skill as software-concept-architect-refine, provides its description, declares English as the language, records its translation source, and sets version 1.0.0.
Main instructions Markdown describes the refinement principles, evidence-gathering process, decision-making, implementation boundaries, and verification expectations.
references/moves.md Required supporting reference for moves and evidence; consult it before selecting a design move.
Analysis evidence Requirements, specifications, relevant code, scenarios, and counterexamples are inputs to the investigation; the skill asks the agent to distinguish facts from inferences.
Generated artifacts No specific output files, persistent data format, or metadata taxonomy for project artifacts is prescribed.

Software Concept Architect · Refine Advanced Features

  • Evidence-led diagnosis uses real scenarios and counterexamples instead of relying on abstract design preferences.
  • Explicitly compares split/merge, unify/specialize, and tighten/loosen options against the root cause, while retaining no change as an option.
  • Makes trade-offs transparent by explaining benefits, costs, and scope of impact.
  • Handles consequential decisions in rounds and waits until prerequisites are settled before discussion.
  • Respects concept contracts when changing synchronization relationships.
  • Separates preserved behavior from intentional changes during authorized implementation.
  • Validates both individual concept purposes and end-to-end scenarios, and calls out anything not verified.

SKILL.md


Loading

Related Openclaw Skills

METADATA

Github Stars: 0
forks: 0

Featured*