Skip to content
Back to skills

Clarify

ASecurity

Advance Cladding onboarding after the user answers a pending product question. Use the MCP prepare/apply flow, preserve the answer verbatim, and never invent answers. Activate only when the connected project contains spec.yaml or the user explicitly names Cladding; ignore ordinary requests in uninitialized projects.

  • 16 stars
  • 0 votes
  • 0 copies
  • 3 views
  • Added September 5, 2026
ai-agentsshell

Works with

  • mcp

Security analysis

A100/100

Scanned September 5, 2026

npx -y skills add qwerfunch/cladding --skill clarify --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Clarify?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Clarify
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/qwerfunch-clarify/badge)](https://www.skillsdirectory.com/skills/qwerfunch-clarify)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
description: Advance Cladding onboarding after the user answers a pending product question. Use the MCP prepare/apply flow, preserve the answer verbatim, and never invent answers. Activate only when the connected project contains spec.yaml or the user explicitly names Cladding; ignore ordinary requests in uninitialized projects.
---

# Cladding clarify

Use this workflow only for an answer contained in a new user message after Cladding displayed the pending question. Never infer, synthesize, or reuse an answer from an initialization turn. If no new user answer exists, stop without calling either clarify tool.

Use this workflow only after Cladding initialization returned a pending question and the user has answered it.

1. Call `clad_prepare_clarify` with the user's answer verbatim.
2. Read the returned prompt, current state, and artifacts. Draft the structured refinement required by `clad_clarify` using the current host model.
3. Call `clad_clarify` with the same answer, the one-time token, and the draft.
4. Ask `nextQuestion` verbatim when one remains.
5. If the result is `needs_review`, show the proposal diffs for every `pendingReview` target and wait for explicit user approval. Only then call `clad_resolve_onboarding_review` with the approved targets. Never describe onboarding as complete while review remains.
6. Report completion only when the returned status is `done`.

Do not run `clad clarify` in a shell from an AI host. Do not use MCP sampling. Never answer a product question on the user's behalf. A stale, malformed, replayed, or answer-mismatched apply request is a no-op and must be prepared again.

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.