Comprehensive Claude Code feature usage: CLAUDE.md as "constitution," subagents, hooks

Source type	Practitioner
Confidence	Medium
Last verified	2026-03-20
Topics	claude-code

Insight: From a team consuming billions of tokens/month for codegen, key CLAUDE.md practices: start with guardrails not a manual (document based on what Claude gets wrong); don't @-file docs (bloats context — instead "pitch" the agent on why and when to read a file); treat CLAUDE.md ad space with token budgets per tool section. The 13KB professional CLAUDE.md only documents tools used by 30%+ of engineers.

Detail: Anti-pattern: embedding entire doc files via @-mentions inflates every run's context. Better approach: mention path + explain when to read it ("For complex usage or if you encounter FooBarError, see path/to/docs.md"). For monorepos, allocate token budgets per section like advertising space — if a team can't explain their tool concisely, it's not ready for the CLAUDE.md. The article serves as a comprehensive reference covering slash commands, subagents, hooks, and GitHub Actions integration.

Sources

• Shrivu Shankar — "How I Use Every Claude Code Feature" (2025-11-02)

Related: existing entry "AI coding quality is a skill issue" in external/claude-code.md — COMPLEMENTS