| name | my-claude-toolkit |
| description | The user's portable, self-authored Claude Code setup — reusable skills, hooks, helper command-scripts, a sanitized MCP-server inventory, and a usage-monitor. Use when setting up Claude Code on a NEW machine, restoring the user's personal toolkit, re-adding their hooks/skills to settings.json, or reviewing which skills/subagents/commands they use most. Not a marketplace plugin: everything here was hand-written by the user and is safe to copy. |
My Claude Toolkit
A single bundle of the user-authored Claude Code artifacts found on their
PC, packaged so the same setup can be reinstalled on another machine without
re-deriving it. Vendor/marketplace plugins (vercel:*, telegram:*,
israeli-*, ralph-loop, skill-creator, etc.) are intentionally excluded
— those reinstall from their marketplaces.
What's inside
| Folder | Contents |
|---|
skills/ | User-written skills not already top-level in this repo (subset-worktree). |
commands/scripts/ | Reusable PowerShell helper scripts (commit-and-pr, verify). |
hooks/ | Hook scripts + hooks.md (event/matcher/command to paste into settings.json). |
mcp/mcp.md | Sanitized MCP-server inventory (names, purpose, required env vars — placeholders only). |
monitor/ | Usage monitor: a SessionEnd/Stop hook + aggregator + one-off history analyzer, monitor/README.md to install. |
USAGE.md | Most-used ranking (skills / slash commands / subagents / MCP) derived from local transcripts. |
README.md | Overview + provenance (where each item came from). |
Already-imported elsewhere in this repo (referenced, not duplicated):
handoff skill → top-level skills/handoff/.- The 13-agent example dev roster →
skills/dev-rules-and-agents/agents/.
Reinstall on a new machine
- Skills — copy each folder under
skills/ to ~/.claude/skills/<name>/
(or into a project's .claude/skills/). subset-worktree's engine script
is invoked as ~/.claude/skills/subset-worktree/New-SubsetWorktree.ps1.
- Command scripts — drop
commands/scripts/*.ps1 into a project's
.claude/scripts/ and run them with
powershell -ExecutionPolicy Bypass -File .\.claude\scripts\<name>.ps1.
- Hooks — copy
hooks/*.ps1 to the target location, then add the matching
hook block from hooks/hooks.md to ~/.claude/settings.json (global) or a
project .claude/settings.json.
- Usage monitor — copy
monitor/*.ps1 to ~/.claude/scripts/, then add
the Stop (and optional SessionEnd) hook from monitor/README.md.
- MCP — recreate servers from
mcp/mcp.md. Set every ${env:NAME} as a
Windows User environment variable first; never hardcode a secret.
Provenance & safety
Every item lists its typical source path in README.md. All secrets (auth JWTs,
DB passwords, bot tokens, connection strings) are stripped and replaced with
${env:NAME} / <REDACTED> placeholders. The real, filled config stays
local — this folder is a public, sanitized template. See mcp/mcp.md and
hooks/hooks.md for the redaction notes.
