
from thx-boris10
Collection of production-tested patterns, templates and protocols for running Claude Code effectively: CLAUDE.md, subagents, hooks, sessions and parallel orches
A practical handbook of patterns used by the Claude Code team to set up, run and scale Claude-based workflows. Includes templates for CLAUDE.md, subagent design, hook automation, session and permission management, and patterns for running parallel Claude sessions.
Use this skill when you need to: bootstrap Claude Code for a project, create focused subagents, add pre/post hooks, manage permissions and sessions, or orchestrate multiple Claude instances for a complex task. Good for teams standardizing Claude workflows.
Likely compatible with Claude Code CLI and Claude web agents (Opus/Sonnet family); requires shell access for included scripts.
Comprehensive Claude Code patterns skill covering CLAUDE.md setup, subagent design, hooks, permissions, parallel sessions, and MCP integration. setup-project.sh ran successfully and created proper config files. The two validate scripts failed only because they ran in an empty temp dir without target files — they're designed to validate existing projects, not run in isolation. Well-organized with clear progressive disclosure via references/ directory.
Clean, well-structured skill. No security concerns. Scripts follow bash best practices (set -euo pipefail, proper heredocs). SKILL.md is well-organized with clear trigger mapping table and step-by-step instructions. The validate scripts failing in isolation is expected behavior, not a defect.