Apollo Prospector

職業

分類

實驗室工具

Expected Apollo actions (actual tool slugs discovered at runtime):

• People search — search for contacts by title, seniority, company size, industry, location
• Organization search — look up company details
• Contact enrichment — get email / phone data

Field Mapping (Apollo response → lead_record schema)

Use jq to transform the raw Apollo response into the lead_record schema:

# Example jq transform for a single Apollo person record
jq '{
  lead_id: .id,
  first_name: .first_name,
  last_name: .last_name,
  title: .title,
  seniority: .seniority,
  department: .departments[0],
  company_name: .organization.name,
  company_domain: .organization.website_url,
  company_industry: .organization.industry,
  company_size: (.organization.estimated_num_employees | tostring),
  company_revenue: (.organization.annual_revenue | tostring),
  company_location: [.organization.city, .organization.state, .organization.country] | map(select(. != null)) | join(", "),
  company_linkedin_url: .organization.linkedin_url,
  linkedin_url: .linkedin_url,
  email: .email,
  email_status: .email_status,
  phone: .phone_numbers[0].raw_number,
  technologies: [.organization.current_technologies[].name],
  source: "apollo",
  pulled_at: (now | todate)
}'

Data Operations

Use jq for all JSON read/write/transform operations:

# Read ICP
jq '.' "$CAMPAIGN_DIR/icp_schema.json"

# Save query params
echo "$QUERY_PARAMS_JSON" | jq '.' > "$CAMPAIGN_DIR/apollo_query_params.json"

# Transform and save leads
echo "$RAW_RESPONSE" | jq '[.people[] | {lead_id: .id, ...}]' > "$CAMPAIGN_DIR/calibration_leads.json"

# Deduplicate by lead_id
jq -s '.[0] + [.[1][] | select(.lead_id as $id | [.[0][] | .lead_id] | index($id) | not)]' \
  calibration_leads.json bulk_leads.json > raw_leads.json

# Update pipeline state
jq '.completed_steps += [3] | .current_step = 3 | .credits_used.apollo += $credits' \
  --argjson credits "$CREDITS_USED" "$CAMPAIGN_DIR/pipeline_state.json" > tmp.json && mv tmp.json "$CAMPAIGN_DIR/pipeline_state.json"

Process

Phase 1: Calibration Pull

1. Load ICP. jq '.' "$CAMPAIGN_DIR/icp_schema.json"

2. Build search parameters. Translate ICP criteria into Apollo search parameters:

   • personas[].title_patterns → person title search
   • personas[].seniority → seniority filter
   • industries → industry filter
   • employee_range → employee count range
   • geo.countries → location filter
   • technologies → technology filter
   • Apply exclusions

3. Save query params. Write apollo_query_params.json via jq for audit trail.

4. Execute calibration pull. Use COMPOSIO_SEARCH_TOOLS to discover the Apollo people search tool, then call it via COMPOSIO_MULTI_EXECUTE_TOOL with limit set to the calibration sample size (typically 50).

5. Transform and save. Use jq to map the response to lead_record schema and write to calibration_leads.json.

6. Report and hand off. Use jq to compute summary stats (title distribution, company distribution) and tell the user. Hand off to human-calibration. Do not proceed to the bulk pull until calibration_feedback.json exists with approved_for_bulk_pull: true.

Phase 2: Bulk Pull (after calibration feedback)

1. Check for feedback. jq '.approved_for_bulk_pull' "$CAMPAIGN_DIR/calibration_feedback.json" — if not true, stop.

2. Adjust search. Use jq to merge feedback adjustments into query params.

3. Execute bulk pull. Call the Apollo people search tool via COMPOSIO_MULTI_EXECUTE_TOOL with the full lead_volume_target limit. Paginate if needed.

4. Deduplicate. Use jq to remove leads already in calibration_leads.json (match on lead_id).

5. Save raw leads. Use jq to merge and write raw_leads.json.

6. Update pipeline state. Use jq to update pipeline_state.json — increment Apollo credits, mark step 3 complete.

7. Hand off to the lead-scorer skill.

Apollo API Reference

See reference/apollo_api_reference.md for endpoint details, parameters, and rate limits.

Output Schema

Each lead in raw_leads.json must conform to reference/lead_record.schema.json (bundled with this skill).