Customizing Your Agent and Repository
FocusCLAUDE.md and AGENTS.md, what to put where; hooks for lint gates, test runs, and guardrails; subagent patterns (planner / implementer / reviewer).
You buildRepository instructions, two hooks, and a planner / implementer / reviewer split
Core material≈ 3 h 40 min
- 1Article30 minClaude Code best practicesAnthropic
The canonical guide to
CLAUDE.md, tool allowlists, workflows, and multi-Claude patterns. - 2Article15 minSteering Claude Code: skills, hooks, rules, subagents, and moreAnthropic
When to use each mechanism.
- 3
- 4Video28 minMastering Claude Code in 30 minutesAnthropic
Boris Cherny's own walkthrough of the configuration surface.
- 5Video50 minInside Claude Code With Its CreatorBoris Cherny
Terminal design,
CLAUDE.md, teams, subagents, plan mode, and the future of coding.
Tools and references
- LinkHooks referenceClaude Code docs
And Create custom subagents — the exact event names, matchers, and frontmatter you need for the build.
- LinkHow to write a great agents.md: lessons from over 2,500 repositoriesRepository instruction files: GitHub
; Real Python, How to Write an AGENTS.md File; Agentic AI Foundation, Writing an Effective AGENTS.md.
- LinkDon't Build Multi-AgentsThe subagent debate: Cognition
Versus Anthropic, How we built our multi-agent research system. Read both before designing your three roles.
- LinkClaude Code in ActionFree courses: Anthropic Academy
; DeepLearning.AI, Claude Code: A Highly Agentic Coding Assistant.
Additional video track
- Full workflow1 h 37 minAI Coding Workflow: From Product Idea to Tested ImplementationMatt Pocock
Human alignment, PRDs, vertical slices, unattended agents, QA, review, and parallelization.
- Engineers' perspective1 h 10 minThe Secrets of Claude Code From the Engineers Who Built ItEvery
, and the Latent Space interview Claude Code: Anthropic's Agent in Your Terminal with Boris Cherny and Cat Wu.
- Hooks in practice30 minI'm HOOKED on Claude Code Hooks: Advanced Agentic CodingIndyDevDan
Pre/post tool hooks, logging, and stopping a destructive command.
- Advanced configurationClaude Code Advanced Patterns: Subagents, MCP, and Scaling to Real CodebasesAnthropic
Hooks, orchestration, guardrails, internal tools, and context strategies.
- Practitioner perspective5 h 15 min; watch the programming-with-agents, vibe-coding-versus-engineering, agent setup, model, and harness chapters onlyFuture of Programming, AI, Modern Software Engineering, Vibe Coding and LinuxDHH with Lex Fridman
. The timestamped transcript makes selective viewing easy.
Build
Add a concise repository instruction file. Treat it as a map, not an encyclopedia: architecture entry points, commands, conventions, safety boundaries, and links to deeper docs. Add one deterministic hook that runs a fast lint or test gate, and one that blocks a dangerous command.
Create three roles (planner, implementer, and independent reviewer) and give each a narrow contract. Run the same feature once with a single agent and once with the three-role workflow. Compare elapsed time, tokens, defects found, and amount of human intervention.