From sdd-workflow
Analyze specifications and generate clarification questions to eliminate ambiguity before implementation
How this skill is triggered — by the user, by Claude, or both
Slash command
/sdd-workflow:clarifyThis skill is limited to the following tools:
The summary Claude sees in its skill listing — used to decide when to auto-load this skill
Scans specifications across 9 key categories and generates targeted clarification questions to eliminate ambiguity
Scans specifications across 9 key categories and generates targeted clarification questions to eliminate ambiguity before implementation.
Read the following prerequisite references before execution:
references/prerequisites_plugin_update.md - Check for plugin updatesreferences/prerequisites_principles.md - Read AI-SDD principles documentreferences/prerequisites_directory_paths.md - Resolve directory paths using SDD_* environment variablesThis command is complementary to the vibe-detector skill:
| Tool | Purpose | When to Use |
|---|---|---|
| vibe-detector | Detects vague instructions in user requests | During task initiation |
| clarify | Scans existing specs for ambiguity and gaps | Before implementation planning |
Output templates are located under templates/${SDD_LANG:-en}/ within this skill directory.
The SDD_LANG environment variable determines the language (default: en).
$ARGUMENTS
| Argument | Required | Description |
|---|---|---|
feature-name | Yes | Target feature name or path (e.g., user-auth, auth/user-login) |
--interactive | - | Interactive mode: Answer questions one at a time |
/clarify user-auth
/clarify task-management
/clarify auth/user-login # For hierarchical structure
Both flat and hierarchical structures are supported.
For flat structure:
Load ${CLAUDE_PROJECT_DIR}/${SDD_REQUIREMENT_PATH}/{feature-name}.md (PRD, if exists)
Load ${CLAUDE_PROJECT_DIR}/${SDD_SPECIFICATION_PATH}/{feature-name}_spec.md (if exists)
Load ${CLAUDE_PROJECT_DIR}/${SDD_SPECIFICATION_PATH}/{feature-name}_design.md (if exists)
For hierarchical structure (when argument contains /):
Load ${CLAUDE_PROJECT_DIR}/${SDD_REQUIREMENT_PATH}/{parent-feature}/index.md (parent feature PRD, if exists)
Load ${CLAUDE_PROJECT_DIR}/${SDD_REQUIREMENT_PATH}/{parent-feature}/{feature-name}.md (child feature PRD, if exists)
Load ${CLAUDE_PROJECT_DIR}/${SDD_SPECIFICATION_PATH}/{parent-feature}/index_spec.md (parent feature spec, if exists)
Load ${CLAUDE_PROJECT_DIR}/${SDD_SPECIFICATION_PATH}/{parent-feature}/{feature-name}_spec.md (child feature spec, if exists)
Load ${CLAUDE_PROJECT_DIR}/${SDD_SPECIFICATION_PATH}/{parent-feature}/index_design.md (parent feature design, if exists)
Load ${CLAUDE_PROJECT_DIR}/${SDD_SPECIFICATION_PATH}/{parent-feature}/{feature-name}_design.md (child feature design, if exists)
Note the difference in naming conventions:
index.md, {feature-name}.md)_spec or _design suffix required (index_spec.md, {feature-name}_spec.md)Analyze specifications across 9 key categories. See references/nine_category_analysis.md for:
Based on category analysis, generate up to 5 high-impact questions using the format in
references/nine_category_analysis.md.
After receiving user answers, the main agent (this skill) applies the integration:
clarification-assistant agent output*_spec.md or *_design.md using Edit/Write toolsNote: The clarification-assistant agent outputs integration proposals (target file, section, content). This skill
applies the actual edits.
Use the templates/${SDD_LANG:-en}/clarification_output.md template for output formatting.
When user provides answers, use --integrate flag:
/clarify user-auth --integrate
This will:
| Scenario | Recommended Action |
|---|---|
| Before task breakdown | Run /clarify to catch ambiguities early |
| After receiving vague requirements | Use with /generate-spec to build complete specs |
| During spec review | Verify all 9 categories are addressed |
| Before implementation | Final check to ensure no hidden assumptions |
/clarify {feature} # Identify ambiguities
|
(Update specs with answers)
|
/check-spec {feature} # Verify consistency
|
/task-breakdown {feature} # Generate tasks
/clarify user-auth --categories flow,integrations,edge-cases
/clarify user-auth --detail minimal # Top 3 questions only
/clarify user-auth --detail standard # Top 5 questions (default)
/clarify user-auth --detail comprehensive # All identified issues
The following verifications are automatically performed during clarification:
# Re-scan to verify clarity improvements
/clarify {feature-name}
# Consistency check (verify updated specifications)
/check-spec {feature-name} --full
| Clarity Score | Recommended Action |
|---|---|
| 80% or above | Ready for implementation |
| 60-79% | Recommended to resolve Partial items |
| Below 60% | Further clarification required before implementation |
npx claudepluginhub toshikiimagawa/ai-sdd-workflow --plugin sdd-workflowAsks up to 5 targeted clarification questions to identify underspecified areas in a feature spec, then encodes answers back into the spec file.
Hunts ambiguous, under-specified points in a spec.md and resolves or defers each one so two engineers can't build different things from the same spec.
Converts vague feature ideas into written, agreed specifications through structured questioning across nine dimensions. Produces docs/specs/*.md with acceptance criteria for planning, scaffolding, and TDD tools.