Optimized Claude Code workflow kit: session startup + handoff system, project documentation skills, outcome-grading loops, spec/fix/prd workflows, and specialist agents (architect, code-verifier, mechanic). Pair with the claude-setup.py bootstrapper for hooks, permission guardrails, and the ccx launcher.
Read-only architecture and design analysis on the strongest available model. Use for planning non-trivial changes, evaluating approaches, or reviewing system design BEFORE implementation. Do not use for writing code.
Adversarial verification of completed work on a strong model. Use after implementing a feature or fix, before claiming done. Tries to REFUTE the claim that the work is correct and complete.
Cheap fast model for mechanical, low-judgment work - batch renames, applying a precisely specified edit across many files, generating boilerplate from an exact template, formatting sweeps. The spec must be complete; this agent must never design or decide.
Detailed commit message format and rules. Use when creating git commits.
Use when tests are failing and you need autonomous fix-test-fix iteration until all tests pass. Trigger on "fix all tests", "make tests green", "fix the test suite", or when multiple test failures need systematic resolution.
Structured bug fix workflow. Creates tasks/fix-[bug].md as a persistent artifact then invokes systematic debugging.
Outcome-based grading loop - define a rubric, have a fresh-context grader evaluate the work against it, and revise until it passes. Use when the user says "grade this", "grade it against...", or wants quality enforced against explicit criteria before delivery.
Interview the user relentlessly about a plan or design until reaching shared understanding, resolving each branch of the decision tree. Use when user wants to stress-test a plan, get grilled on their design, or mentions "grill me".
Uses power tools
Uses Bash, Write, or Edit tools
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
Save tokens, reduce costs, and get better results from Claude Code.
Based on official Anthropic documentation, community benchmarks, and real-world testing. These practices can reduce your Claude Code usage by 50-70%.
Step-by-step guide designed for non-technical users. Takes about 10 minutes.
Open index.html in your browser or view online: coach-foundation.github.io/claude-code-best-practices
The wizard walks you through:
New to Claude Code or want optimized defaults? One-command setup (backs up your existing files):
git clone https://github.com/Coach-Foundation/claude-code-best-practices.git
cd claude-code-best-practices
python3 claude-setup.py || python claude-setup.py
Then use cc instead of claude to start.
Have existing settings you want to keep? Claude analyzes your setup and merges best practices without overwriting customizations:
git clone https://github.com/Coach-Foundation/claude-code-best-practices.git
cd claude-code-best-practices/smart-optimizer
claude
Claude will automatically walk you through the recommendations.
Open claude-code-best-practices.html in your browser for a visual, non-technical guide. Save as PDF with Cmd+P / Ctrl+P.
Or view it online: coach-foundation.github.io/claude-code-best-practices/claude-code-best-practices.html
| Setting | What it does | Impact |
|---|---|---|
| MAX_THINKING_TOKENS=10000 | Caps expensive thinking tokens | ~70% thinking cost reduction |
| Subagent model: Haiku | Uses cheaper model for helper tasks | ~80% cheaper subagents |
| Effort level: medium | Reduces output verbosity | 50-70% fewer output tokens |
| Auto-compact: 75% | Triggers compaction earlier | More buffer, fewer emergencies |
| Context Efficiency rules | Stops Claude from echoing, narrating, re-reading | Major output reduction |
| .claudeignore | Blocks scanning of junk files | Prevents context blowups |
After setup, start a Claude Code session and install these:
/install-plugin superpowers # Structured workflows (planning, TDD, code review)
/install-plugin context7 # Always-current library documentation
/install-plugin code-simplifier # Code quality reviews
Found a tip that saves tokens? Open a PR or issue.
MIT
npx claudepluginhub coach-foundation/claude-code-best-practices --plugin ap-optimal-claudeUpstash Context7 MCP server for up-to-date documentation lookup. Pull version-specific documentation and code examples directly from source repositories into your LLM context.
Comprehensive feature development workflow with specialized agents for codebase exploration, architecture design, and quality review
Comprehensive C4 architecture documentation workflow with bottom-up code analysis, component synthesis, container mapping, and context diagram generation
A growing collection of Claude-compatible academic workflow bundles. Covers scientific figures, manuscript writing and polishing, reviewer assessment, citation retrieval, data availability, paper reading, literature search, response letters, paper-to-PPTX conversion, and evidence-grounded Chinese invention patent drafting. Rules are organized as reusable skill folders with explicit workflows and quality checks.
Harness-native ECC operator layer - 67 agents, 271 skills, 92 legacy command shims, reusable hooks, rules, selective install profiles, and production-ready workflows for Claude Code, Codex, OpenCode, Cursor, and related agent harnesses
Upstash Context7 MCP server for up-to-date documentation lookup. Pull version-specific documentation and code examples directly from source repositories into your LLM context.