The command execution capability serves the CLI’s development assistant purpose, but one-shot Claude use silently bypasses the approval prompt and permits model-generated shell commands. This is a real consent risk, so the package warrants a warning rather than a publish block.
Static reason
One or more suspicious static signals were detected.; previous stored version diff introduced dangerous source
Trigger
Run `smbls ask` with prompt arguments while Claude is the configured provider.
Impact
Model-generated commands can read or change project files and perform other shell actions without a separate confirmation.
Mechanism
The CLI switches to automatic permission mode and executes Claude tool requests through a shell command in the current working directory.
Evidence
dist/esm/bin/ask.jsFiles accessible from process.cwd() through shell commands
Network endpoints1
api.anthropic.com/v1/messages