A collection of skills, hooks, and utilities for Claude Code CLI.
| Name | Description | Status |
|---|---|---|
| setup-sudo-askpass | Set up graphical sudo password prompt for GUI-based authorization | ✅ Ready |
| mobile-game-test | Drive a Moto G32 via Appium on the lab VM (192.168.1.19) to test the AI-GM player UI end-to-end on real Android. Bundles a MobileSession Python helper with 39 methods covering auth, hero/campaign creation, turn loop, dice prompt detection, combat, loot, and modals. |
✅ Ready |
| taskmaster-github-setup | Set up TaskMaster CLI with GitHub Issues sync for any project — creates tasks from a PRD, links each to a GitHub issue, and establishes the "add task" workflow. | ✅ Ready |
| code-cleanup | Safely remove dead code, duplicate logic, unused components, and unnecessary complexity — one reviewed batch at a time. Hybrid engine (static tools → adversarial agent → runtime tracing), user approves every deletion, 7 protective layers (pytest + smoke + log tracing + 14-day quarantine + git isolation). Never drops DB tables, never deletes without approval. | ✅ Ready |
| tdd | Test-Driven Development for AI-GM with plain-Polish narration. Full Red→Green→Refactor; builds a pytest test AND a Playwright spec per issue, runs only that issue's tests on DEV, updates the GitHub issue. | ✅ Ready |
| workflow-loop | Turn an agreed spec into a "one-prompt work system": per-task description doc + checklist (single source of truth) + reusable starting prompt. Each fresh session runs exactly ONE task end-to-end then STOPs with a plain-Polish report. Modeled on AI-GM's "FAZA U". | ✅ Ready |
| game-smoke | Full-mode smoke playtest of AI-GM: ~15 real LLM turns through a complete mechanics scenario (movement, NPC, quest, combat, shop, rest+XP, beats), each checkpoint verified against the DB. Deliverable: checkpoint table + P0/P1/P2 defect issues. | ✅ Ready |
| game-test-player | Claude impersonates a player to reproduce/verify a specific issue end-to-end: creates a #NNN campaign on Demo, plays 8–12 turns, runs a DB check, posts a text report to GitHub. Bundles setup scripts. |
✅ Ready |
| game-test-player-screenshot | Same as game-test-player, plus screenshots at key moments embedded in the GitHub comment for visual evidence. | ✅ Ready |
| playwright-test-report | Drive a web app through a requested test scenario via the Playwright MCP browser, screenshot each step in order, file full TDD-style issues for bugs, and produce a plain-language step-by-step report (screenshots + "what you see / pass or problem") ending with a priority-ordered fix list linking every issue. Headless; runs on a display-less VM. | ✅ Ready |
| Name | Description | Status |
|---|---|---|
| Coming soon |
| Name | Description | Status |
|---|---|---|
| Coming soon |
See individual skill/hook README files for installation instructions.
# Clone the repo
git clone git@github.com:szmidtpiotr/ClaudeCode-Tools.git ~/ClaudeCode-Tools
# Create Claude skills directory
mkdir -p ~/.claude/skills
# Symlink each skill *directory* into ~/.claude/skills/ so bundled
# resources (scripts/, references/) come along, not just SKILL.md.
for skill_dir in ~/ClaudeCode-Tools/Skills/*/; do
name="$(basename "$skill_dir")"
ln -sfn "$skill_dir" "$HOME/.claude/skills/$name"
done- auto-git-commit - Hook to auto-suggest commit messages
- setup-dev-env - Skill to bootstrap common development environments
- backup-claude-config - Utility to backup/restore Claude Code settings
- project-init - Skill to initialize new projects with CLAUDE.md templates
This is a personal tools repository. Feel free to fork for your own use.
MIT