Do NOT use when a prompt contains $key or $key=args or $key::section tokens — in Claude Code those are already resolved by the UserPromptSubmit hook and injected as context before the message arrives. Use when the user explicitly says to use or fetch a rune directly (e.g., "use the docs rune", "run crunes", "fetch rune output"), or when you want to inspect rune output mid-conversation, or when the agent needs live context before exploring, planning, or touching code. Prefer targeted calls over broad ones unless scope is wide or unknown.
To fetch the output of a rune manually:
crunes use <key-token> [-a <key-token>]...
key[=arg1,arg2][::section1,section2]
| Separator | Meaning |
|---|---|
= | Args (comma-separated) |
:: | Section filter (comma-separated) |
: | Plugin key separator only (e.g. my-plugin:runeKey) |
Examples:
crunes use docs — fetch all sections from the docs runecrunes use api=v2 — pass v2 as an argument to the api runecrunes use api=v2::endpoints — pass arg and only return the endpoints sectioncrunes use api::endpoints,errors — section filter with no argscrunes use api=v2 -a git -a env::vars — batch: three runes in one callUse -a to use multiple runes in a single invocation:
crunes use structure -a git -a env
crunes use api=v2::endpoints -a git::commits --format json
All sections are returned in order, combined into one output.
crunes use docs # markdown (default)
crunes use docs --format json # JSON Section[] array
crunes -p use docs # plain output (no ANSI)
In prompts, tokens are resolved automatically by the UserPromptSubmit hook:
$key all sections
$key=arg1,arg2 with args
$key::section section filter
$key=arg1::section1,section2 args + section filter
my-plugin:runeKey=arg plugin rune with args
Use this skill when the user explicitly asks to inspect rune output, or when the agent needs live context before exploring, planning, or touching code. Prefer targeted calls — fetch only what the task needs.
crunes use api -a missing --fail-fast # stop on first error
crunes use api -a missing # default: run all, exit 1 if any failed