Skip to main content

Print the LLM Syntax Reference (mdslide llms)

The llms command prints the complete SYNTAX.md reference — the canonical, authoritative specification of mdslide's Markdown syntax, layouts, annotations, and CLI commands — as plain markdown to stdout. It's built for piping straight into an AI coding assistant's context window, or saving to a file for later reference.

Usage​

mdslide llms

Examples​

# Print the full syntax reference to the terminal
mdslide llms

# Save it to a file so an AI agent can read it as context
mdslide llms > SYNTAX.md

Options & Flags Reference​

llms takes no command-specific flags. Only the shared Global Flags are accepted, and in practice only --timeout has any real effect on this command — the output is a fixed, already-compiled markdown string, so --json, --no-input, --yes, and --dry-run are accepted (they won't be rejected) but change nothing about what gets printed.

FlagTypeDefault ValueDescription
--timeout <ms>numbernoneAbort the command after <ms> milliseconds (exit code 124). The only global flag with a real effect here.

There is no --json mode for this command: the output is already plain, machine-readable markdown, so there is nothing additional a JSON envelope would add — llms always writes the same reference text to stdout regardless of flags passed.

Why This Command Exists​

Modern AI coding assistants write better mdslide decks when the exact syntax rules — layout auto-detection heuristics, comment annotation syntax, frontmatter keys, CLI flags, and error codes — are loaded directly into their context instead of guessed at or hallucinated from partial examples. mdslide llms exists to make that trivial:

  1. No network access required. The reference is bundled into the CLI package itself at build time, so mdslide llms works completely offline and returns instantly — no fetching docs from a website.
  2. Always matches the installed version. Because the text is bundled with the CLI binary, the syntax reference you get always matches the exact behavior of the mdslide version installed, rather than documentation that may have drifted from an older or newer release.
  3. Typical agent workflow: run mdslide llms > SYNTAX.md once at the start of a session and add the resulting file to the AI assistant's context (or pipe it directly into a prompt). From then on, the agent knows the exact rules for slide separators, layouts (title, bullets, split, code, quote, visual, table, statement), comment annotations, admonitions, charts, image overrides, speaker notes, and every CLI command and flag — without needing to open a browser or read the Docusaurus site.
  4. Plain, undecorated output. Unlike other commands, llms never applies ANSI colors or spinners — the output is always the raw markdown text, safe to redirect straight into a file or another tool.

If you're building an AI-agent-driven workflow around mdslide, running mdslide llms > SYNTAX.md and referencing that file is the fastest way to make sure the agent's understanding of the syntax stays current with the installed CLI.