clank

Fork, clone to ~/.clank, run installer, edit CLAUDE.md

RED-GREEN-REFACTOR for process documentation - baseline without skill, write addressing failures, iterate closing loopholes

Skills wiki intro - mandatory workflows, search tool, brainstorming triggers

Interactive idea refinement using Socratic method to develop fully-formed designs

Execute detailed plans in batches with review checkpoints

Execute implementation plan by dispatching fresh subagent for each task, with code review between tasks

Complete feature development with structured options for merge, PR, or cleanup

Create isolated git worktrees with smart directory selection and safety verification

Receive and act on code review feedback with technical rigor, not performative agreement or blind implementation

Search previous Claude Code conversations for facts, patterns, decisions, and context using semantic or text search

Create detailed implementation plans with bite-sized tasks for engineers with zero codebase context

Four-phase debugging framework that ensures root cause investigation before attempting fixes. Never jump to solutions.

Run verification commands and confirm output before claiming success

TDD for process documentation - test with subagents before writing, iterate until bulletproof

Write the test first, watch it fail, write minimal code to pass

Never test mock behavior. Never add test-only methods to production classes. Understand dependencies before mocking.

Dispatch code-reviewer subagent to review implementation against plan or requirements before proceeding

Hide implementation details behind interfaces - work at domain level (what), not implementation level (how)

Class interfaces present one cohesive abstraction - don't mix domain logic with serialization, persistence, or unrelated concerns

Comment WHY code exists and non-obvious decisions, not WHAT code does (mechanics)

Try 2-3 different approaches before implementing - don't settle for first design you think of

Each routine does one thing and does it well - extract when routines have multiple responsibilities

Declare variables in smallest possible scope, initialize close to first use, minimize span and live time

Name code by what it does in the domain, not how it's implemented or its history

Choose names that fully and accurately describe what the variable represents

Refactor with tests first, one change at a time, never mix refactoring with bug fixes or new features

Flatten nested conditionals with early returns or table-driven methods - keep nesting depth under 3 levels

Use each variable for exactly one purpose - no hybrid coupling or hidden meanings

Check all external inputs for validity - garbage in, nothing out, never garbage out

Write comments explaining WHAT and WHY, never temporal context or history

Maintain skills wiki health - check links, naming, cross-references, and coverage

Managing complexity is software's primary technical imperative - all other goals are secondary

Design in pseudocode first, iterate approaches, then translate to code

Use multiple Claude agents to investigate and fix independent problems concurrently

Validate at every layer data passes through to make bugs impossible

Systematically trace bugs backward through call stack to find original trigger

Replace arbitrary timeouts with condition polling for reliable async tests