Your task
- •
Check if the user provided a project name (e.g.
/gumbo:init myapp). If they did, pass it as the third argument to the init script. If not, omit the third argument and the script will default to the directory name. - •
Run the gumbo init script:
# With project name: ./scripts/init.sh ~/.gumbo "$PWD" <project-name> # Without project name (defaults to directory name): ./scripts/init.sh ~/.gumbo "$PWD"
- •
Ensure
CLAUDE.local.mdexists in the project root ($PWD/CLAUDE.local.md). If it doesn't exist, create it. - •
Ensure
CLAUDE.local.mdcontains a reference to.gumbo/AGENTS.local.md. If the file already has this reference, leave it alone. Otherwise, add the following line:
@.gumbo/AGENTS.local.md
- •
Check the project's
.gitignoreforCLAUDE.local.mdandAGENTS.local.md. If either is missing from.gitignore, add it. These are local/private files that shouldn't be committed to the code repo. - •
After successful initialization, remind the user to add the gumbo data directory to their Claude Code context so it's available across sessions. Tell them to run:
/add-dir ~/.gumbo
They can add it to either the current project or to their user-scoped settings (~/.claude/settings.json) if they want it available in all projects.
Report the output to the user. Do not use any other tools or do anything else.