🧰 CLAUDE.md

The items that name it 7

• TokenMaxxing: How to Get the Most Out of Your AI Models
  article2026-10-17

  Instruction file the agent reads at the start of every session; keep under 200 lines (also called Soul.md)

• How the Claude Code team does code review
  article2026-10-16

  Instruction file telling Claude what the project is (stack, conventions, goals, non-goals); loaded on every review; kept short and pointing to other docs

• Getting the most out of Opus 5.5 in Claude and Claude Code
  article2026-10-03

  A project instructions file used to set standing rules for Claude Code, such as when to stop and ask versus keep going, or the format of end-of-run summaries

• Claude Code’s Next Era — Thariq Shihipar, Anthropic
  talk2026-09-29

  Markdown file used to give Claude Code persistent instructions/goals and record repeated failure modes; Thariq expects it may eventually disappear as models improve.

• Fighting AI Slop in Production Codebases: How Huntress Improved Fable 5.1’s API Recall
  writeup2026-09-22

  Agent context file where the API recall rule (and lookup tool description) were added as instructions to the coding agent

• Using Claude’s Security Review Skill in Claude Code
  article2026-07-29

  Project instructions file Claude Code reads for project-specific context and conventions

• 7 Ways to Reduce Claude Code Token Usage
  article2026-05-13

  A memory file that loads before Claude reads code or the task and persists in the context window for the entire session (never lazy-loaded or evicted); used to store stable project instructions like how to run tests, package manager, formatting rules, and directories to avoid.

In every workspace you can read 12