From skill-reviewer
Guides users through planning, designing, and scaffolding new Claude Code skills from scratch. Use when the user asks to "plan a skill", "design a skill", "create a new skill", "scaffold a skill", "help me build a skill", or "what design pattern should my skill use". Covers design pattern selection, folder structure, frontmatter drafting, and file generation — not skill review or auditing.
npx claudepluginhub shawn-sandy/agentics --plugin skill-reviewerThis skill uses the workspace's default tool permissions.
Walks users through a structured workflow to plan and generate a complete Claude Code skill. Produces a ready-to-use skill folder with SKILL.md, optional reference files, and optional scripts.
Writes articles, guides, blog posts, tutorials, and newsletters in a voice from examples or brand guidance. For polished long-form content with structure, pacing, and credibility.
Records polished WebM UI demo videos of web apps using Playwright with cursor overlay, natural pacing, and three-phase scripting. Activates for demo, walkthrough, screen recording, or tutorial requests.
Executes repo commands, inspects git state, debugs CI failures, and pushes narrow fixes with exact proof of execution and verification. Use for command runs, repo checks, or evidence-based changes.
Walks users through a structured workflow to plan and generate a complete Claude Code skill. Produces a ready-to-use skill folder with SKILL.md, optional reference files, and optional scripts.
Follow these steps exactly.
Before doing any other work, use TodoWrite to create todos for each step of this workflow. This gives the user visibility into progress.
Create the following todos (all starting with status: "pending"):
Mark each todo status: "completed" as you finish that step.
Ask the user up to 4 questions using AskUserQuestion to understand what they want to build. Tailor questions to what the user has already provided — skip questions they have already answered. If the user's initial message already covers purpose, triggers, tools, and expected output, skip directly to the concept summary without asking questions.
Questions to consider (pick the most relevant):
After gathering answers, summarize the skill concept:
**Skill concept:**
- Purpose: [what it does]
- Trigger: [when it activates]
- Tools/services: [what it uses]
- Output: [what the user gets]
Based on the skill concept from Step 1, recommend a design pattern. Present the options using AskUserQuestion with descriptions from references/design-patterns.md.
If the skill concept clearly fits one pattern, recommend it as the first option with "(Recommended)" in the label. If ambiguous, present 2–3 patterns and let the user choose.
After selection, confirm:
**Design pattern:** [Pattern name]
**Why:** [One sentence explaining the fit]
Based on the skill concept and design pattern, determine what the skill folder needs.
Always include:
SKILL.md — core instructionsInclude references/ when:
Include scripts/ when:
Include assets/ when:
Present the proposed structure to the user:
my-skill/
├── SKILL.md
├── references/
│ └── [planned-reference-files]
└── scripts/
└── [planned-scripts]
Ask if they want to adjust the structure before proceeding.
Generate the name and description fields following these rules:
Name rules:
anthropic or claude as substringreviewing-code) over imperative (review-code)Description rules:
Present the drafted frontmatter:
---
name: [generated-name]
description: [generated-description]
---
Ask the user to confirm or adjust before proceeding.
Generate a body outline based on the design pattern selected in Step 2. Each pattern has a recommended structure — see references/design-patterns.md for details.
All patterns include:
Body quality targets:
If the outline suggests the body will exceed 3,000 words, recommend creating a reference file for the most detailed section before proceeding. If the skill serves two distinct user intents, recommend a Skill Pack (multiple skills in one plugin) instead of a single oversized skill.
Present the outline as a numbered list of sections with brief descriptions:
1. ## Overview — What the skill does, freedom level
2. ## Step 1 — [First action]
3. ## Step 2 — [Second action]
...
Ask the user to confirm or adjust the outline.
After the user confirms the outline, generate the complete skill files on disk.
Generation order:
SKILL.md with:
Before writing, confirm the target directory:
"I'll create the skill at
[path]/[skill-name]/. Should I proceed?"
If the user declines or wants a different location, ask: "Where would you like me to create the skill folder? Please provide the path."
After generation, present a summary:
## Skill Generated
**Location:** `[path]/[skill-name]/`
**Files created:**
- `SKILL.md` ([line count] lines, [word count] words)
- `references/[file]` (if any)
- `scripts/[file]` (if any)
**Next steps:**
1. Review the generated SKILL.md and adjust as needed
2. Test activation by asking Claude something that matches the trigger phrase
3. Run the `reviewing-skills` skill to audit quality