Daily driver

One week of day-2 Free CLI prompts — health, watch, investigate, optimize, scale/rollback plans, session — shipped capabilities only (IC-015).

Muscle memory for the Free CLI after Day-0. Swap payments / api for your namespace and workload. Prefer a sandbox context first. Reads run immediately; mutates stop on a PlanResult unless you say y / --approve.

The week

  • Day 1 — Health pulse: kprompt "how's my cluster" (no LLM key)
  • Day 2 — Laptop scan: kprompt watch -n payments --once (suggestions only; never applies)
  • Day 3 — Root cause: kprompt "investigate api" -n payments (or why is api crashing)
  • Day 4 — Rightsizing read: kprompt "optimize my cluster" (report only)
  • Day 5 — Scale plan: kprompt "scale api to 3" -n payments → review → --approve only if trusted
  • Day 6 — Rollback plan: kprompt "rollback api" -n payments → same gate
  • Day 7 — Habit close: kprompt session · history · optional remember

Copy-paste

# Day 1–2 — observe
kprompt "how's my cluster"
kprompt watch -n payments --once

# Day 3–4 — reason (read-only)
kprompt "investigate api" -n payments
kprompt "optimize my cluster"

# Day 5–6 — plan, then maybe approve
kprompt "scale api to 3" -n payments
# kprompt "scale api to 3" -n payments --approve
kprompt "rollback api" -n payments

# Day 7 — local memory
kprompt session
kprompt history
kprompt remember "payments = Team A"

Honesty

  • Do run Days 1–4 on production if RBAC allows reads
  • Do not --approve Days 5–6 against prod on first try
  • Escape hatch: kprompt demo ($0 magic moment)
  • MCP / IDE is for plan review — apply stays CLI
  • Graduate to recipe packs when the week sticks

Related: Demo ($0) · MCP · Adopt.