How this skill is triggered — by the user, by Claude, or both
Slash command
/magus:magusopusThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
The `magus` CLI program provides an agentic coding workflow that:
The magus CLI program provides an agentic coding workflow that:
All of magus' files are saved to ./magus/plans/, ./magus/memories/ and ./magus/skills.
Magus supports the following command-line arguments:
| Flag | Description |
|---|---|
-p, --prompt <file> | Read prompt from a file instead of stdin |
--resume <id> | Resume a previous session by its ID |
--auto-approve | Skip interactive plan approval |
-v, --verbose | Show full plan details with dependencies |
-H, --hide-tools | Suppress tool |
IMPORTANT: After producing a plan, Magus' default behaviour is to use the TTY to ask the user if they approve the plan or have feedback.
Because we can not interact with the TTY easily from Claude Code, Magus must always be run from Claude Code with the --auto-approve flag to
trigger automatic execution of the coding phase.
Remember: You can learn magus' full set of supported command-line arguments by running magus -h or magus --help.
The following rules are MANDATORY when working with Magus and ensure the smoothest experience for the user possible.
Rules:
Makefile available. Magus can only execute make commands.The following guidelines are best practices that you SHOULD follow to facilitate the best user experience.
Guidelines:
-v or --verbose flag.-H or --hide-tools flag.Note that the -H and --hide-tools flag DO NOT hide edit tool calls, which display diffs of the changes made to files.
If you are ever uncertain and the user would want you to confirm their choices, use the AskUserQuestion tool to do so. Only ever ask about these optional flags.
Makefile exists. If it does not, explain its necessity to the user and offer to create one with simple "test" and "lint" commands.magus -p <path/to/plan.md> --auto-approve [optional flags].magus is running in a task and monitor the task for completion.Provides behavioral guidelines to reduce common LLM coding mistakes, focusing on simplicity, surgical changes, assumption surfacing, and verifiable success criteria.
Searches, retrieves, and installs Agent Skills from prompts.chat registry using MCP tools like search_skills and get_skill. Activates for finding skills, browsing catalogs, or extending Claude.
Creates, edits, and optimizes skills for Claude Code, including drafting, evaluating with test prompts, iterating on performance, and improving skill descriptions for better triggering accuracy.
npx claudepluginhub atelier-arcadia/magus --plugin magus