What is Prompt Shaping in Claude Code?

Prompt Shaping is a metaskill that helps AI agents interpret underspecified requests by restating intent, structuring the plan, and defining verification loops.

Does Prompt Shaping help prevent errors?

Yes, by making assumptions explicit and requiring a verification step, it ensures the agent doesn't proceed based on a misunderstanding of your intent.

When should I use the Prompt Shaping skill?

Use it whenever a request is vague or incomplete, allowing the agent to make reasonable assumptions and verify them with you before acting.

How does the grounding loop work?

It defines success criteria and a verification method (like a unit test) that the agent must execute and pass before the task is considered complete.

Can Prompt Shaping integrate with existing test suites?

Absolutely. The skill specifically suggests using existing patterns and test suites as part of its grounding and verification process.

Prompt Shaping

Name: Prompt Shaping
Author: lherron

bylherron

生産性とワークフロー

Clarifies underspecified user requests by proposing structured interpretations and verifiable next steps to ensure alignment before execution.

概要

Prompt Shaping is a specialized skill designed to bridge the gap between vague user intent and concrete technical implementation within AI agent harnesses. By restating requests in natural language, breaking them into scannable structural components, and defining a rigorous 'grounding loop' for verification, this skill prevents wasted compute and incorrect code changes. It utilizes environment-aware context grounding to infer reasonable defaults and proposes specific success criteria—such as automated tests or command assertions—to guarantee that the agent's actions fulfill the user's underlying requirements accurately.

主な機能

Actionable Roadmaps: Provides a clear 'next step' commitment to move from planning to execution.
Grounding Loops: Establishes clear success criteria and verification methods before starting work.
Iterative Verification: Commits to running tests or commands until the proposed solution passes.
Intent Prototyping: Rapidly clarifies ambiguous requests through structured proposals.
Context Awareness: Scans existing project patterns to make informed, reasonable assumptions.
0 GitHub stars

ユースケース

Creating a 'test-driven' development plan for bug fixes by proposing a failing test case as a success criterion.
Defining the boundaries of complex refactoring tasks where the user hasn't specified which modules should be affected.
Clarifying high-level feature requests like 'add caching' by making assumptions about scope and invalidation strategies explicit.

概要

主な機能

Actionable Roadmaps: Provides a clear 'next step' commitment to move from planning to execution.
Grounding Loops: Establishes clear success criteria and verification methods before starting work.
Iterative Verification: Commits to running tests or commands until the proposed solution passes.
Intent Prototyping: Rapidly clarifies ambiguous requests through structured proposals.
Context Awareness: Scans existing project patterns to make informed, reasonable assumptions.
0 GitHub stars

ユースケース

Creating a 'test-driven' development plan for bug fixes by proposing a failing test case as a success criterion.
Defining the boundaries of complex refactoring tasks where the user hasn't specified which modules should be affected.
Clarifying high-level feature requests like 'add caching' by making assumptions about scope and invalidation strategies explicit.