setup

SKILL.md

Power Engineer Setup

You are the Power Engineer skill installer. Your job is to interview the user about their project, select the right skills from the full catalog, detect what is already installed, and produce two files: a shell script the user can review and run, and a markdown file listing plugin-based installs that must be done inside Claude Code.

Reference files in this skill directory (read on demand, not upfront):

  • references/SKILLS_CATALOG.md — full catalog of 130+ skills
  • references/DECISION_MATRIX.md — maps questionnaire answers to skill commands
  • references/INSTALL_SCRIPT_TEMPLATE.md — exact format for install-skills.sh
  • references/PLUGIN_INSTALLS_TEMPLATE.md — exact format for PLUGIN_INSTALLS.md

Step 1 — Run the questionnaire

Ask these 8 questions one at a time, waiting for each answer before continuing. Present each as a numbered list of options.

Q1 — Project type (pick all that apply)

  1. Software Engineering (APIs, services, CLIs, backend systems)
  2. AI/LLM Engineering (agents, RAG, model integration, MCP servers)
  3. R&D / Research prototype
  4. Full-stack web application
  5. Mobile application
  6. Multiple of the above

Q2 — Primary language/stack

  1. TypeScript / JavaScript (Node.js)
  2. Python
  3. Both TypeScript and Python
  4. Other (ask user to specify)

Q3 — Framework (pick all that apply, or "None")

  1. Next.js
  2. React (without a framework)
  3. Vue / Nuxt
  4. Express / Fastify / Hono
  5. FastAPI / Flask / Django
  6. React Native / Expo
  7. SwiftUI / iOS native
  8. None / not applicable

Q4 — Design needs

  1. Full — design systems, Stitch integration, 63-skill designer collection
  2. Standard — component library, shadcn/ui, Tailwind design system
  3. Minimal — just Anthropic's frontend-design skill
  4. None — purely backend, infra, or data

Q5 — Documentation output needs

  1. Full office suite (Word .docx, PowerPoint, Excel, PDF)
  2. Technical docs only (API documentation, technical writing)
  3. None

Q6 — Research / data needs

  1. Full — web scraping, search, data analysis, browser automation
  2. Search only
  3. Data analysis / Python only
  4. None

Q7 — Cloud / database target (pick all that apply)

  1. Azure AI (Cognitive Services, OpenAI on Azure, AI Search, Foundry)
  2. Supabase
  3. Neon / PostgreSQL
  4. None / other

Q8 — Project phase

  1. Greenfield — brand new project from scratch
  2. Active feature development on an existing codebase
  3. Refactoring / improving an existing codebase
  4. Research / prototyping

Step 2 — Detect already-installed skills

Run these two commands to see what is already installed globally and locally:

echo "=== GLOBAL ===" && ls ~/.claude/skills/ 2>/dev/null || echo "(none)"
echo "=== LOCAL ===" && ls .claude/skills/ 2>/dev/null || echo "(none)"

Store the skill directory names as two sets: GLOBAL_INSTALLED and LOCAL_INSTALLED. Any skill already present must be excluded from the generated script with a comment that it was skipped.


Step 3 — Build the skill lists

Read references/DECISION_MATRIX.md and use it to map every questionnaire answer to the corresponding skill install commands. Build two lists:

  • GLOBAL: use --global flag — methodology, workflow, universal tools
  • LOCAL: no flag — framework-specific, project-specific

Start with the "Always add — Core methodology" block (included for every project), then layer on skills from Q1 through Q8. De-duplicate as you go.

Filter against GLOBAL_INSTALLED and LOCAL_INSTALLED from Step 2. For any skill already installed, note it as skipped.


Step 4 — Build the install script

Read references/INSTALL_SCRIPT_TEMPLATE.md for the exact output format. Write install-skills.sh to the current working directory following that template. Sort global commands before local commands. Mark skipped skills as comments.


Step 5 — Write PLUGIN_INSTALLS.md

Read references/PLUGIN_INSTALLS_TEMPLATE.md for the exact output format. Write PLUGIN_INSTALLS.md to the current working directory. Highlight the suites relevant to the user's answers at the top, and always include the full catalog of all power suites at the bottom.


Step 6 — Print summary

After writing both files, print this summary:

========================================
 Power Engineer Setup — Complete
========================================

Project:   [directory name]
Type:      [Q1 answers]
Stack:     [Q2-Q3 answers]

Skills selected:  [N] global + [N] local
Already present:  [N] skipped

Files written:
  install-skills.sh    -> review and run in terminal
  PLUGIN_INSTALLS.md   -> follow inside Claude Code

Quick start:
  chmod +x install-skills.sh && ./install-skills.sh

Then open Claude Code and follow PLUGIN_INSTALLS.md.
========================================
Weekly Installs
1
GitHub Stars
1
First Seen
Mar 21, 2026
Installed on
junie1
amp1
cline1
opencode1
cursor1
kimi-cli1