Three practical Claude Code skills born from real-world production use:
multi-ai-agents — Distribute tasks across Claude, Gemini CLI, and Codex CLI in parallel. Type x3, x5, or x7 to spawn multi-agent workloads where Claude orchestrates, Gemini analyzes, and Codex generates code. Includes quality gate and synthesis step.
typefix — Systematic TypeScript type error fixer. Runs tsc --noEmit, categorizes errors by type and file, fixes them with minimal targeted changes (never as any or @ts-ignore), then verifies zero errors remain. Supports scoped fixes via file/folder filter.
kickstart — Structured co-founder planning session for any new project. Claude asks 5 critical questions, then generates a master plan with roadmap, directory structure, knowledge base context files, CLAUDE.md, and mode routing (when to use Deep Research vs Code vs MCP). Everything saved as files, not chat.
multi-ai-agents: Claude Code token costs add up on large tasks. By offloading read-only analysis to Gemini CLI (free) and mechanical code generation to Codex CLI, you get 3x the throughput at lower cost. Claude keeps final decision authority.
typefix: After refactors, dependency upgrades, or AI code generation, you often face dozens of TypeScript errors. Manually fixing them is tedious. This skill automates the detect → categorize → fix → verify loop.
kickstart: Most projects fail in the first hour — jumping into code without structure. This skill forces a planning phase that produces reusable context files, so every future Claude Code session starts with full project understanding.
gemini -p and codex exec🤖 Generated with Claude Code