Shell completion and agent skill

Shell completion

Add one line to your shell’s startup file:

source <(beachside completion zsh)     # ~/.zshrc
source <(beachside completion bash)    # ~/.bashrc

For fish, save the script once:

beachside completion fish > ~/.config/fish/completions/beachside.fish

Tab then completes commands, flags, --provider ids, --profile names from beachside.json, and sandbox names:

beachside exec <tab>          # web  job-1
beachside cp ./app.py <tab>   # web:  job-1:

Sandbox names come from the CLI’s own list, so pressing tab never calls a provider.

Agent skill

The CLI carries an agent skill that teaches coding agents how to use it: check the setup first, always use --json, prefer run for one-off work, read exit codes, never retry an unknown outcome, and clean up. The skill always matches the installed CLI version.

beachside skill install            # for you, in every project
beachside skill install --project  # for this repository only

install writes the skill to .agents/skills/beachside-cli/, read by most agents, and to .claude/skills/beachside-cli/ when Claude Code is set up. Run it again after upgrading the CLI to update the skill. It never replaces a copy you installed another way.

beachside skill prints the skill, for agents without a skill system: paste it into their instructions.

The skill can also be installed with npx skills add pandemicsyn/beachside --skill beachside-cli. See Agent skills for the SDK skill as well.