LinkedIn Hiring Lead Enricher — emails verified + buying committee inferred

You are helping me build a LinkedIn Hiring Lead Enricher agent. This is the enrichment layer that sits between a hiring-signal scraper (which produces partial contacts) and an outbound sequencer. It does two jobs every run:

JOB A — Enrich existing leads. Take rows from `hiring_leads` with `enrichment_status='pending_enrichment'`, fetch their full LinkedIn profile, find their email via the agntdata multi-provider waterfall, verify the email, and update the row.

JOB B — Multi-stakeholder discovery. For job posts where `Get_Hiring_Team` returned nothing (`job_posts.status='no_hiring_team_found'`), fall back to a people-search by company + role-leadership filters. Insert every qualifying stakeholder into `hiring_leads` as new pending_enrichment rows.

The agent runs hourly and processes up to 50 leads in Job A + 10 job posts in Job B per run.

REFERENCE DOCS (read these before writing code)
- Full agntdata API documentation: https://agnt.mintlify.app/apis/overview
- agntdata orchestrator endpoints (used here):
  - `people_email_finder` — multi-provider waterfall (Hunter → Prospeo → LeadMagic etc.); one call, one credit charge.
  - `people_email_verifier` — returns `deliverable` / `risky` / `undeliverable`.
  - `people_search` — company + filters → stakeholders. Better-curated than calling LinkedIn directly.
  - `people_enrich`, `people_bulk_enrich` — single-call enrichment for cases where you have a linkedin_url and want everything (profile + email + verify) in one shot.
- LinkedIn endpoints used as fallbacks: `Get_Profile_Data_By_URL`, `Search_People`.

ABOUT MY MOTION
- Product name: <YOUR PRODUCT>
- One-line description: <WHAT IT DOES>
- Who I sell to: <BUYER ROLE FAMILY — same as the upstream scraper>
- Upstream scraper: <linkedin-hiring-signal-scraper or your own equivalent — produces `hiring_leads` + `job_posts` tables>

ROLE CLASSIFICATION
Job B classifies discovered stakeholders into 4 buckets and inserts only two of them. Define your title→role mapping:

- `decision_maker` — final-call buyer. Default titles: Founder, Co-founder, CEO, President, COO, CRO.
- `function_owner` — functional owner of the area you sell into. Default titles for selling to GTM: VP Sales, Head of Sales, Director of Sales, VP Revenue, CMO, etc. Customize for your buyer.
- `recruiter` — Talent / Recruiting / HR. Inserted only if you sell to recruiters; otherwise SKIPPED.
- `other` — anything else. SKIPPED.

WHAT TO BUILD
- A scheduled agent on agntdata that runs hourly (claude-sonnet-4-6 — role classification benefits from the smarter model).
- Job A: claim pending_enrichment leads atomically (UPDATE to 'in_progress' first), then waterfall through profile fetch → email finder → email verifier. Update the row's enrichment_status to 'verified' / 'risky' / 'invalid' / 'email_not_found'.
- Job B: scan `job_posts.status='no_hiring_team_found'`, call people_search per company with seniority + department filters, classify each result, dedupe against `hiring_leads.linkedin_profile_url`, insert qualifying rows as new pending_enrichment leads. Update the job_post status to 'hiring_team_inferred' on success.
- DO NOT push to Instantly or any external CRM. That is the downstream sequencer's job.

DATABASE (must already exist — owned by the upstream scraper)
- `hiring_leads` — has at minimum: id, linkedin_profile_url, full_name, company_name, title, job_post_id, raw_hiring_team_data, enrichment_status. This blueprint adds: first_name, last_name, company_domain, email, email_verified, email_status, stakeholder_role.
- `job_posts` — read `company_name` and `status`; update `status` to 'hiring_team_inferred' when Job B succeeds.

DELIVERY
- Compact JSON summary at the end of each run with Job A + Job B counts (see WORKFLOW below for the schema).

GUARDRAILS
- Max 50 leads per Job A run, max 10 job posts per Job B run — keep runs bounded.
- Never re-process where `enrichment_status != 'pending_enrichment'`. The atomic UPDATE to 'in_progress' is the lock.
- Wrap each lead in try/catch — on error, set `enrichment_status='error'` and continue.
- Multiple contacts per company is expected and correct. Do not skip a company because another lead from there exists.
- Never push to Instantly here. The sequencer handles activation.

When you're ready, start by asking me the ABOUT MY MOTION and ROLE CLASSIFICATION blocks.