lince

Run and verify code in a disposable, isolated lab VM, and bisect regressions. Use when asked to test a fix in a clean machine, drive the lince quickstart wizard, check install.sh idempotency, reproduce a bug in isolation, or find which commit broke something. Authors a recipe TOML, runs it via the lince-lab CLI (broker over a unix socket), and reads the result artifacts — never runs limactl or touches the host filesystem directly.

Configure LINCE sandbox and dashboard settings in natural language. Use when asked to set up providers, change security levels, modify agent settings, configure API keys, adjust sandbox behavior, or any "how do I configure" question about LINCE. Drives the Config v2 policy layer (~/.config/lince/lince.toml) through the lince-config CLI — apply/set/validate only, never direct TOML edits.

Add support for a new AI coding agent (or CLI tool) to lince-dashboard and agent-sandbox. The agent provides its own requirements (binary, config dirs, API keys, sandbox behavior) and this skill generates correct TOML configuration. Use when adding a new agent type, setting up multi-agent support, or when asked to add agent, new agent type, support cursor, support bob, support cli, register agent, setup agent, configure agent for dashboard.