setup

**end-to-end install wizard:** takes the guesswork out of securing your agentic workspace - probes your machine's state and maps out a detailed roadmap all the way through - interactive questionnaire helps you decide which sandbox config is right for you - ends with a clean `--audit` handoff that makes sure everything is fully secure - runs settings, permissions, scripts, credentials and upstream, then keeps each output whole - recomputes no verdict; each lens grades itself and this collects what they returned - correlates across lenses, which no single lens can do from inside itself - takes no lens flag, since one lens belongs to its own skill and its own artifact - prices the run against the sidecars it would replay, and asks before spending any of it - costs minutes, and names the seconds each lens spent so a long stage reads as work # Instructions ## Telemetry ```! "${CLAUDE_PLUGIN_ROOT}"/skills/setup/setup.sh $ARGUMENTS echo "sidecar exit: $?" ``` - `help: requested` → the run was refused before it started; `## Help` below is the whole turn - `confirm: required` → the suite was priced and nothing ran; `## Confirm` below is the whole turn - it already ran with whatever flags the invocation carried - NO FLAG → the machine was probed; continue to step 1 - `--roadmap` → a reprint ran and nothing was probed; skip to step 3 - `--audit --confirm` → the whole suite ran under `lib/suite.sh`; skip to step 4 - fail (`sidecar exit` > 0) → a preflight refused; report wh...