Generate List

Step 1 — Classify Entity Type

From "$ARGUMENTS", determine whether the user wants:

• Businesses — companies, accounts, organisations (no person role mentioned)
• Prospects — people, contacts, leads, decision-makers (job title or role mentioned)

If unclear, ask one question: "Are you looking for companies or for specific people at those companies?"

Step 2 — Parse ICP into Filters

Extract structured filter values from the natural language description:

Shared filters:

• Geography → country_code / city_region_country (need autocomplete)
• Company size → company_size (exact values: 1-10, 11-50, 51-200, 201-500, 501-1000, 1001-5000, 5001-10000, 10001+)
• Revenue range → company_revenue
• Tech stack → company_tech_stack_tech / company_tech_stack_categories (need autocomplete)
• Intent topics → business_intent_topics (need autocomplete)

Business-only filters:

• Industry → linkedin_category or google_category (always need autocomplete)
• NAICS → naics_category (need autocomplete)

Prospect-only filters:

• Job title → job_title (need autocomplete)
• Seniority → job_level (exact values: cxo, vp, director, manager, senior, entry, owner, partner)
• Department → job_department (exact values: engineering, sales, marketing, finance, product, c-suite, data, human resources, operations, etc.)
• Convenience flags → --has-email, --has-phone

Step 3 — Resolve Autocomplete Filters

For every filter in the "need autocomplete" list, run autocomplete before building the final filter set. Always use --semantic for best matching:

AUTO=$(mktemp /tmp/vibep-auto.XXXXXX.json)

node "$CLI" businesses autocomplete linkedin_category "software development" --semantic \
  --call-reasoning "$QUERY" > "$AUTO"

node "$CLI" prospects autocomplete job_title "head of engineering" --semantic \
  --call-reasoning "$QUERY" > "$AUTO"

Extract the exact value strings from the response and use them verbatim in filters. Never guess or paraphrase autocomplete values.

Step 4 — Size the Market (Free)

Run stats before any fetch. This is always free and prevents unexpected large fetches:

node "$CLI" businesses stats \
  --filters '{"linkedin_category":{"values":["Software Development"]},"company_size":{"values":["51-200","201-500"]}}' \
  --call-reasoning "$QUERY"

node "$CLI" prospects stats \
  --filter job_level=vp,director \
  --filter company_size=51-200 \
  --has-email \
  --call-reasoning "$QUERY"

Show the user the total result count. If the count is very large (>10,000), suggest narrowing the filters before fetching.

Step 5 — Preview a Sample

Fetch 10 rows in preview mode to show the data shape before committing to a larger pull:

PREVIEW=$(mktemp /tmp/vibep-preview.XXXXXX.json)

node "$CLI" businesses fetch \
  --filters @filters.json \
  --limit 10 \
  --mode preview \
  --call-reasoning "$QUERY" > "$PREVIEW"

node "$CLI" prospects fetch \
  --filter job_level=vp,director \
  --filter company_size=51-200 \
  --has-email \
  --limit 10 \
  --mode preview \
  --call-reasoning "$QUERY" > "$PREVIEW"

Present the sample as a table:

Businesses:

Prospects:

Then ask the user:

"Found [total] matching [businesses/prospects]. Showing 10 of those. How many would you like to fetch? Fetching uses credits. Suggest a limit or say 'all' for the full set."

Do not proceed to full fetch without explicit confirmation.

Step 6 — Full Fetch

Once the user confirms a volume, fetch the full set. For large requests, write to file to keep context clean:

RESULTS=$(mktemp /tmp/vibep-results.XXXXXX.json)

node "$CLI" businesses fetch \
  --filters @filters.json \
  --limit 500 \
  --call-reasoning "$QUERY" > "$RESULTS"

node "$CLI" prospects fetch \
  --filter job_level=vp,director \
  --filter company_size=51-200 \
  --has-email \
  --limit 500 \
  --call-reasoning "$QUERY" > "$RESULTS"

Inspect only metadata back into context to confirm the fetch completed:

node -e "
const d = JSON.parse(require('fs').readFileSync(process.argv[1], 'utf8'));
console.log('total_results:', d.meta.total_results);
console.log('returned:', d.data.length);
console.log('sample:', JSON.stringify(d.data.slice(0,2), null, 2));
" "$RESULTS"

Step 7 — Present Summary

Show a brief summary of the completed fetch:

List generated: [ICP summary]

Step 8 — Offer Next Actions

Ask the user:

1. Export to CSV — Re-run with --to-file results.csv or convert the JSON file
2. Research specific items — Run /vibeprospecting-plugin:research on selected rows for deep dossiers
3. Enrich the list — Run enrich on the results for firmographics, tech stack, contacts, or intent signals
4. Fetch events — Run events on the results to surface trigger signals (funding, hiring, role changes)
5. Refine filters — Adjust the ICP and re-run stats + fetch