Claude Lessons: practice Claude Code and AI workflow skills
Claude Lessons teaches task framing, source grounding, planning, verification, recovery,
and reusable Claude Code workflows through short lessons, role paths, and copyable prompts.
38 Claude Code lessons
- Send a context packet: define goal, audience, constraints, source, and output.
- Make Claude inspect before acting: ask it to understand files before editing.
- Point Claude at the right source: avoid guessing from memory.
- Use plan mode before edits: make the plan visible before files change.
- Define done with acceptance criteria: make completion measurable.
- Ask for the smallest useful step: avoid overbuilding.
- Let Claude interview you: answer key questions before execution.
- Make Claude verify its work: ask for passed, failed, and unverified items.
- Read evidence, not reassurance: request specific files, commands, and gaps.
- Recover when the agent drifts: pause, summarize state, and redirect safely.
- Create and clean memory with /init, /memory, and CLAUDE.md.
- Use permission modes and /permissions deliberately.
- Manage context with /context, /compact, /clear, checkpoints, /rewind, and /resume.
- Review diffs with /diff, /review, and /code-review before trusting changes.
- Extend Claude Code with skills, hooks, MCP, plugins, subagents, and agent teams.
- Use CLI automation, auto mode, output styles, and living playbooks safely.
Enable JavaScript to use the simulated Claude Code workspace and save progress.