Installs into .claude/skills of the current project.
Are you the author of Spec Clarification?
Add the live security badge to your README. It updates with every re-scan.
[](https://www.skillsdirectory.com/skills/b4san-spec-clarification)
---
name: spec-clarification
description: Identify underspecified areas in the current feature spec by asking targeted clarification questions and encoding answers back into the spec.
---
# Specification Clarification
## Overview
Detect and reduce ambiguity or missing decision points in the active feature specification and record the clarifications directly in the spec file.
**Goal**: Identify 5 actionable clarification questions that materially impact implementation.
## Execution Steps
1. **Initialize Context**:
* Load the current spec file (`spec.md`).
* Perform a structured ambiguity & coverage scan.
* Identify missing or partial information in: Functional Scope, Domain & Data Model, Interaction & UX Flow, Non-Functional Quality, Integration & Dependencies, Edge Cases, Constraints, Terminology, Completion Signals, Misc/Placeholders.
2. **Generate Prioritized Questions**:
* Create a queue of candidate clarification questions (max 5).
* **Constraints**:
* Answerable with Multiple Choice (2-5 options) OR Short Answer (<=5 words).
* Materially impacts architecture, data modeling, task decomposition, test design, UX, or compliance.
* Exclude style preferences or plan-level details unless blocking.
* Favor impact * uncertainty heuristic.
3. **Sequential Questioning Loop (Interactive)**:
* Present **ONE** question at a time.
* **Multiple Choice**:
* Analyze options based on best practices/risk/alignment.
* Present **Recommended Option** with reasoning first.
* Show table of options (A, B, C...).
* Prompt user to reply with letter or "yes"/"recommended".
* **Short Answer**:
* Provide **Suggested Answer** based on context.
* Prompt user to reply "yes"/"suggested" or own answer.
* **After Answer**:
* Validate answer.
* Record in working memory.
* Stop if: All critical ambiguities resolved, User says stop, or 5 questions reached.
4. **Integration (Incremental Update)**:
* **After EACH accepted answer**:
* Ensure a `## Clarifications` section exists (create if missing).
* Append a bullet: `- Q: <question> → A: <final answer>`.
* **Apply the clarification** to the appropriate section(s) in the spec:
* Functional → Functional Requirements.
* User Interaction → User Stories/Actors.
* Data → Data Model.
* Non-Functional → NFRs.
* Edge Case → Edge Cases/Error Handling.
* Terminology → Normalize terms.
* **Save the spec file** immediately.
5. **Validation**:
* Ensure exactly one bullet per answer.
* Ensure no contradictory earlier statements remain.
* Ensure Markdown structure is valid.
6. **Report Completion**:
* Summarize questions asked & answered.
* List sections touched.
* Show coverage summary table (Resolved/Deferred/Clear/Outstanding).
* Recommend next steps (e.g., Proceed to Planning).
## Principles
* If no ambiguities, state "No critical ambiguities detected".
* Never exceed 5 questions.
* Avoid speculative tech stack questions unless blocking functional clarity.
* Respect explicit user stops.