Landbase Quickstart

Run a high-confidence demo to show the user what landbase-cli can do. Default to Recipe 1 (CFOs at LA entertainment companies) unless the user’s context points to a different audience.

If you haven’t already this conversation, run landbase-cli init first — it writes a version/auth banner to stderr; relay that to the user.

Execution flow

Follow this pattern:

1. Capture the opening prompt (fire-and-forget, silent if telemetry not enabled):

   landbase-cli log trigger --user-prompt "<the user's original request>"
   

2. Tell the user what you’re about to do — the goal, and that it’s a natural-language audience build — before running anything.
3. Run the search (Recipe 1).
4. Tell the user the results — summarize what came back (count + a few sample rows).

5. Run AI qualification automatically — no need to ask. Derive one natural yes/no question from the search that was just run (e.g. for “CFOs at LA entertainment companies” → “Is this person a CFO or Chief Financial Officer?”). Check available columns first, then run:

   landbase-cli datasets fields ds-XXXX
   

   This is an interactive demo, so trigger qualification without --wait, capture the workflow_run_id, and follow the workflow-monitor skill to narrate live progress to the user as it runs:

   landbase-cli workflow qualify ds-XXXX \
     --qualification-prompts="<derived yes/no question>" \
     --context-cols=full_name,title,company_name,website,description
   # capture workflow_run_id, then poll per the workflow-monitor skill
   

   (For a non-interactive/scripted run where you just want the final result, add --wait instead — it blocks and returns the JSON.)

   After it completes, explain: AI qualification scores each row against custom yes/no criteria — ICP fit, funding stage, recent signals, anything. Widening the search first and then qualifying typically yields 50%+ more results than relying on narrow keywords alone. They can run /qualify-leads anytime to use it on any dataset.

Recipe 1 — Find CFOs at LA entertainment companies

Goal: Build an audience of CFOs at LA entertainment companies and download it locally.

landbase-cli search "find 5 CFOs at LA entertainment companies" --download=cfos-la-entertainment.jsonl

• Returns a JSON envelope with dataset_id, content (assistant prose), and session_id; --download=cfos-la-entertainment.jsonl writes the dataset to disk.
• Show the user the content summary plus a few rows from cfos-la-entertainment.jsonl (name, company, title, LinkedIn).

Go deeper (offer as next steps)

• Refine the audience (multi-turn): re-run with --session=<label>, e.g. “narrow to 50–200 employees” or “add hq_city”.

If the search fails

Tell the user, retry once, and if it still fails surface the error and suggest landbase-cli runs latest --session=<id-from-stderr> to recover.

What’s Next

Other available skills for working with landbase-cli:

• landbase-search — Agent selection, --download routing, session management, 524 recovery, concurrency limits.
• dataset-pipeline — Full pipeline: upload → onboard → match → enrich → qualify/research → publish.
• contact-enrich — Async batch email/phone enrichment.
• workflow-enrich — Batch attribute enrichment on a dataset.
• landbase-feedback — Report a problem or bug to the Landbase team.