`gtmesh discover`

Installs the project-discovery skill into an empty directory and prints a prompt you paste into Claude Code — the bootstrap that runs before init.


gtmesh discover "what you sell, audience, market, URL"

What it does

gtmesh discover is LLM-free: it installs the discovery skill into .claude/skills/ and hands you a prompt. You open the directory in Claude Code and paste that prompt. The discovery skill then interviews you, pulls a demand-validated information architecture (sections, clusters, the page types it needs, harvest classes, and clustered starter terms), writes gtmesh-discovery.yaml, and stops.

That file then feeds gtmesh init --from-discovery, so your scaffold starts tailored to your business instead of the generic Acme demo.

The CLI only installs the skill and prints the prompt. The interview, the demand pulls, and the file write all happen inside Claude Code — that is where the model lives.

Arguments & flags

Argument	Description
`[description...]`	Optional business description woven into the printed prompt

There are no command-specific options. The global --project <dir> selects the target directory (default .).

Reads & writes

Reads (CLI): nothing of yours; it copies the bundled discovery skill.
Writes (CLI): .claude/skills/ (the installed skill) and the prompt to stdout.
Writes (the skill, in Claude Code): gtmesh-discovery.yaml — which you review and edit before init.

Examples


mkdir mybrand && cd mybrand
gtmesh discover "we sell water pumps to UK contractors; site at example.com"
#   → installs the discovery skill and prints a prompt to paste into Claude Code

gtmesh discover