A new open-source tool introduces 'executable Markdown' by combining Unix pipes with AI coding assistants like Claude Code. The core innovation is a shebang (`#!/usr/bin/env claude-run`) that allows Markdown files to be executed like shell scripts. These files act as dynamic prompts that can invoke tool use—running shell commands, reading files, and making API calls—replacing complex Python orchestration code with readable text files.

The workflow supports standard Unix piping, enabling commands like `cat data.json | ./analyze.md` to process data through AI. A key benefit highlighted is improved auditability; unlike opaque `curl | bash` install scripts, users can read the Markdown source to verify actions before execution. This approach aims to make AI-driven automation composable, shareable, and transparent, potentially changing how developers build and audit automated workflows.