Yeda AI Knowledge Base

Claude Code

Getting more out of Anthropic’s coding agent: context, plans, rules, and subagents.

Español

← All categories

Tip #011

Plan Before You Prompt: Review the Plan Before Your AI Agent Touches Code

Put your coding agent in plan mode before it edits a single file. One keystroke gets you a reviewable, editable plan instead of a surprise diff.

Tip #012

Keep Your Rules File Lean

Your CLAUDE.md gets loaded into every request — a bloated rules file wastes context and dilutes instructions. Here’s what belongs in it and what doesn’t.

Tip #013

Give Grunt Work Its Own Agent: Subagents for Context Isolation

How Claude Code subagents isolate context — delegate noisy research to a helper agent with its own window so only the summary lands in your main thread.

Tip #014

Make It Ask First: The 95%-Confidence Prompt

Stop revision loops before they start — tell Claude Code to ask clarifying questions until it’s 95% confident, then push it to name what you’re missing.

Tip #015

Keep Your Context Fresh: Fighting Context Rot in Claude Code

Long Claude Code sessions quietly lose quality — context rot. Learn when to /clear, when to /compact with keep-notes, and how to watch your context percentage.

Tip #209

The Rule Your Agent Keeps Ignoring

When an AI coding agent repeatedly breaks a rule you wrote down, the cause is usually file length, not weak wording. Here’s the diagnostic and the fix.

Tip #210

Your Helper Agent Never Read Your Rules

A delegated subagent starts with an empty context and may never receive your project rules at all. What actually loads, what doesn’t, and when to fork instead.

Tip #211

Editing Your Rules Mid-Session Does Nothing

Rules files are read once at session start and held in memory, so a mid-session edit is invisible until you restart, clear, or compact. Why, and the three ways to reload.

Tip #212

Your History Is Deleted In 30 Days

Claude Code stores session transcripts locally and sweeps them after cleanupPeriodDays — 30 days by default, with no warning. How to change it and what’s exempt.

Tip #213

Parallel Agents Cost Seven Times More

Agent teams use roughly 7x the tokens of a standard session in plan mode. That multiplier applies to teammates, not to summarizing subagents, which are the context-saving option.

Tip #214

The Cheaper Model That Costs More

Prompt caching is a per-model prefix match, so switching models mid-conversation reprocesses your whole context at full price — twice, if you switch back.

Tip #215

Stop Being The Verification Loop

If nothing checks your agent’s work, you are the check and every error waits for a human. Four escalating verification layers, and how to scope a reviewer so it doesn’t invent findings.

Tip #216

The Commands Nobody Types

A tour of built-in Claude Code commands most people never open: /doctor, /insights, /branch, /btw and /subtask — what each does and when it’s actually the right move.

Tip #217

Your Agents Can’t Talk Without Names

Claude Code sessions can message each other, but only a named session is addressable. How to name them, list what’s reachable, and where messages actually travel.

Tip #218

Another Agent’s Message Is Not Your Approval

A message from one Claude Code session to another can’t grant consent, change configuration, or execute a command. The trust boundary between your agents, and how to tighten it.

Tip #219

Your Subagent Didn’t Vanish

A finished subagent can be resumed with its full history — every tool call, result and piece of reasoning — instead of spawning a blank replacement. How resuming works and why naming matters.

Tip #220

Your Agent Can’t Test What You Never Gave It

Agents stop at unit tests because their reach stops there — no browser, no CLI, no throwaway credentials. Ask what it needs to verify end to end, then build that once.

Tip #221

Delete Your Scaffolding For One Task

Most AI-coding setups are fossils built for what models needed a year ago, and nobody re-tests them. Run one real task with your rules and skills moved aside, and compare.

Tip #222

Your Best Skill Never Runs

Skills are selected from their description, not their content — and when you have many, the least-used descriptions get trimmed from the listing first, making rare skills unfindable.