Anthropic's official "Code with Claude" workshop materials — nine hands-on courses on skills, MCP, evals and Claude Managed Agents.
- official
- workshops
- anthropic
- managed-agents
Claude Code Tips
55 field-tested Claude Code tips across nine sections — sessions, commands, subagents, skills, MCP, hooks, workflow and context — one short recipe each.
A living collection of the author's favorite Claude Code techniques as he explores the tool, organized into General, Command, Agent, Skill, Mcp, Prompt, Hooks, Workflow and Context sections. The recipes are concrete and short: name and resume sessions like git branches, let /batch run a migration across hundreds of files, run parallel sessions in isolated worktrees, define reusable subagents and have a fresh one grade the work, restrict a skill to manual invocation, tune how aggressively MCP tool definitions load, auto-format every edit with a PostToolUse hook, use a Stop hook as a deterministic gate for unattended runs, keep CLAUDE.md healthy with /doctor, and run risky unsupervised work in a container.
What it does
- 55 tips in 9 sections — General (13), Command (8), Agent (5), Skill (7), Mcp (6), Prompt, Hooks, Workflow and Context
- Session craft: name/resume sessions like branches, move them between laptop/phone/cloud, track spend with ccusage, watch context with /context and /insights
- Hooks and automation: PostToolUse auto-format, a Stop hook as a gate for unattended runs, worktree isolation on non-git VCS, ambient audio cues
- Context discipline: manage context deliberately instead of filling it, keep plans out of the 30-day cache, let /doctor prune CLAUDE.md, reserve IMPORTANT for the one rule Claude keeps missing
When to use it
Daily-driver users who want concrete recipes beyond the official docs — skim the table of contents, adopt one tip a day; also a good syllabus for teams onboarding onto Claude Code.
Getting started
# reference repo — read on GitHub: https://github.com/onmyway133/claude-code-tips-tricks
Things to know
No license file at the time of writing — treat it as reading material rather than something to fork and redistribute. Tips are written against Claude Code as of September 2026; the tool moves fast, so expect details to drift.
Repository
onmyway133/claude-code-tips-tricksCharted
Related on the map
Anthropic's announcement of Fable 5.1 and Mythos 5.1 — one base model, two guardrail postures.
- official
- models
- announcement
- fable
Anthropic's canonical essay on agent design — workflows vs agents, and five composable patterns.
- official
- agents
- engineering
- patterns