feat: scaffold first-party CLI for v2 notebook workflows #5

Summary

Add a first-party MindSpring CLI for agent/operator workflows against the v2 notebook API.

Why

We now have a functional backend route surface for notebook lifecycle, source registration, ingestion jobs, scoped retrieval/chat, and artifacts. A thin CLI will improve reliability for:

• AI agents orchestrating workflows
• operators doing repeatable notebook ingestion tasks
• CI automation and smoke validation

Proposed package

• Package name target: @stackbilt/mindspring-cli (or repo-local tools/mindspring-cli.mjs as phase 0)

Command surface (MVP)

• notebooks create|list|get|update|delete
• sources register|list|get
• jobs status|wait
• search
• chat
• artifacts create|list|get

Config

• MINDSPRING_BASE_URL
• MINDSPRING_API_KEY

Output modes

• human-readable default
• --json machine mode for agent chaining

Acceptance criteria

1. CLI can run the full workflow:
   • create notebook
   • register source upload
   • wait for job completion
   • run notebook chat/search
   • create and fetch artifact
2. --json mode returns stable contracts suitable for automation.
3. Documentation includes install + usage examples.
4. Smoke script can optionally use CLI path instead of raw curl/HTTP.

Suggested implementation plan

• Phase 0: repo-local script in tools/ with strict flags and --json
• Phase 1: harden UX/errors, add tests
• Phase 2: publish npm package and versioning policy