"Extend with llm, stabilise with code" - Prompts when you're exploring, Python when you're certain. Workers and tools form one function space—refactor between them.
10 files for Claude Code, Codex and OpenCode: llm-do, ask-questions-if-underspecified, decomplect-py, coupling-analyzer and 6 more — 1,562 tokens loaded in every session.
AGENTS.md A 4 tok CLAUDE.md A 1,558 tok .claude/settings.json A — .claude/skills/ask-questions-if-underspecified/SKILL.md A 24 tok .claude/skills/decomplect-py/SKILL.md A 55 tok .claude/skills/tasks-system/SKILL.md A — .claude/skills/decomplect-py/commands/decomplect-py.md A 45 tok .claude/skills/decomplect-py/agents/coupling-analyzer.md A 50 tok .claude/skills/decomplect-py/agents/fcis-analyzer.md A 53 tok .claude/skills/decomplect-py/agents/simplicity-analyzer.md A 75 tok These files are zby/llm-do's own configuration — they tell Claude Code, Codex and OpenCode how to work on this repository, so they are not mods to install elsewhere. Copy one as a starting point and replace the parts that are about this project.