prospect-research-to-cadence（前景研究到节奏）

<quick_start> Single prospect: "research prospect Baylor University" → enriches org + decision-makers → drafts personalized outreach → presents for approval → loads to Apollo sequence

Batch mode: "build cadence for these 5 companies: [list]" → runs pipeline for each → presents batch draft → loads approved contacts

Trigger phrases:

• "research prospect [company/domain]"
• "build cadence for [company]"
• "outreach for [company]"
• "research-to-cadence [company]"
• "enrich and sequence [list]"
• "new prospect batch" </quick_start>

<success_criteria>

• Company enriched with firmographics, tech stack, news, and Epiphan use case angle
• 3-5 ATL decision-maker contacts identified per approved title patterns (VP/Director/Dean/Chief/Provost = economic buyers; Managers flagged as GRAY for manual review)
• Personalized 3-touch email sequence drafted per contact
• Call script with MEDDIC discovery hooks generated
• Draft presented for Tim's approval BEFORE any sequence loading
• On approval: contacts created in Apollo + added to specified sequence
• Golden Rules exclusions enforced (no customers, no channel, no product-usage contacts) </success_criteria>

Pipeline Stages

INPUT                    RESEARCH                 DRAFT                    APPROVE              LOAD
─────────────────────────────────────────────────────────────────────────────────────────────────────
Company/Domain    →  Apollo Org Enrich     →  Email Sequences    →  Tim Reviews     →  Apollo Sequence
                  →  Apollo People Search  →  Call Scripts       →  Approves/Edits  →  Contact Create
                  →  Epiphan CRM Check     →  ICP Score          →                  →  Add to Cadence
                  →  Web Research          →  Pain Hypothesis    →                  →

Stage 1: Research (Tools Used)

1a. Company Intelligence

Use these MCP tools in parallel:

1a-bis. Clay Enrichment (Waterfall)

For companies/contacts where Apollo data is incomplete (missing phone, tech stack, or funding data):

Clay acts as a waterfall fallback — it aggregates 50+ data providers. Use it when Apollo returns incomplete results, especially for phone numbers.

1b. Golden Rules Filter

STOP and skip if ANY of these are true:

• lifecyclestage = 'customer' in HubSpot
• first_conversion contains 'Pearl', 'setup', 'Connect', 'signup'
• Company device_count >= 1 in Epiphan CRM
• engagement_overview contains product usage
• is_channel = true
• hubspot_owner_id IN ('82625923', '423155215', '190030668') — AEs: Lex Evans, Ron Epstein, Phillip Sandler → 90-Day Stale Exception: If hs_lastmodifieddate > 90 days ago → SURFACE as STALE AE LEAD with ATL/BTL tier + deal value. Tim reviews to re-engage, push to demo, or help close. → Ron Epstein: all leads. Lex & Phil: NA only. <90 days → still exclude.

OPERATING PRINCIPLE: Tim is NET-NEW first, but can re-engage STALE AE leads (90+ days inactive) to recover pipeline value.

1c. Contact Discovery

Use apollo_mixed_people_api_search with filters:

• Titles (ATL-only): VP, Director, Dean, CIO, CTO, CFO, Provost, Superintendent, Court Administrator, City Manager, Senior Pastor
• Gray Zone Titles (flag for review): Manager (AV/Facilities/IT) — include ONLY if Apollo confirms reports to Director+ AND has delegated budget >$25K
• Exclude (NEVER ATL): Warehouse Manager, Network Manager, Systems Administrator, AV Technician, Graphic Design Instructor, Program Administrator, Web Designer, Classroom Support, Lab Coordinator, Maintenance, Building Engineer, Multimedia Services Manager, Video Production Specialist, Streaming Crew
• Seniority: director, vp, c_suite
• Departments: IT, engineering, operations, education
• Limit: 3-5 contacts per company

For each contact, run apollo_people_match to get verified email + phone.

ATL/BTL Classification (applied to each discovered contact)

Classify each contact using approved title patterns (CLAUDE.md § ATL/BTL Classification v1.0):

• ATL: Chief, VP, Director, Dean, Provost, Superintendent, Court Administrator, City Manager, Senior Pastor, Executive Pastor
• BTL: Technician, Specialist, Coordinator, Support, Administrator (Systems/Network/Database), Engineer (AV/Network/Systems), Operator, Instructor/Professor/Faculty, Designer, Assistant, Clerk (non-Court Admin), Volunteer, Intern, Student, Resident, Help Desk
• NEVER ATL: Warehouse Manager, Network Manager, Systems Administrator, AV Technician, Graphic Design Instructor, Program Administrator, Web Designer, Classroom Support, Lab Coordinator, Maintenance, Building Engineer, Multimedia Services Manager, Video Production Specialist, Streaming Crew
• GRAY: Manager (AV/Facilities/IT) — ATL only if reports to Director+ AND has delegated budget >$25K. Flag for Tim's manual review before sequence load

Tag each contact: atl_btl_tier = 'ATL' | 'BTL' | 'GRAY' Only ATL contacts are auto-approved for sequence loading. GRAY contacts require Tim's explicit approval. BTL contacts are deprioritized (included in batch but called last).

For contacts where apollo_people_match returns no phone number, fall back to Clay's find-and-enrich-contacts-at-company for phone waterfall.

Phone Verification Priority

ALWAYS attempt phone verification. Tim needs 50+ daily dials. Every contact loaded into a sequence MUST have a verified phone number. Priority order: Apollo phone → Clay waterfall → skip if no phone found (flag for manual research).

1d. ICP Scoring

Score the prospect using Tim's ICP verticals:

Boost +5 for: hiring AV roles, recent facility expansion, CMS mentioned in tech stack Boost +10 for: Extron/Matrox replacement signal (competitors exited market)

Stage 2: Draft Outreach

Email Sequence (3-touch)

Load reference/email-templates.md for templates. Customize with:

• Touch 1 (Day 0): Pain-forward, reference specific news/signal, single CTA (15-min call)
• Touch 2 (Day 3): Value prop with peer social proof (similar vertical), ask different question
• Touch 3 (Day 7): Breakup email with resource offer (case study or guide)

Rules:

• Under 100 words per email
• No "I hope this finds you well"
• Lead with THEIR problem, not Epiphan features
• Reference a specific trigger (job posting, news, tech stack gap)
• CTA = specific time ask ("15 minutes this Thursday?")

Call Script

Generate MEDDIC-structured call script:

OPENER (15 sec): Reference email + trigger signal
PAIN PROBE (60 sec): "What's your current setup for [use case]?"
METRICS: "How are you measuring [quality/uptime/cost]?"
ECONOMIC BUYER: "Who else would need to sign off?"
DECISION CRITERIA: "What would you evaluate?"
CHAMPION TEST: "If this made sense, would you be open to a quick demo?"

Deliverable Format

Present draft as structured output:

═══════════════════════════════════════════
PROSPECT RESEARCH: [Company Name]
ICP Score: [XX/100] | Vertical: [vertical]
═══════════════════════════════════════════

COMPANY INTEL:
- Industry: [...]
- Size: [...] employees | Revenue: [...]
- Tech stack: [...]
- Trigger signal: [...]
- Epiphan angle: [...]

CONTACTS (ready for sequence):
┌──────────────────────────────────────────┐
│ 1. [Name] — [Title] [ATL/GRAY/BTL]       │
│    Email: [...] | Phone: [...]           │
│    LinkedIn: [...]                        │
│    Personalization hook: [...]            │
└──────────────────────────────────────────┘

EMAIL SEQUENCE:
Touch 1: [subject] → [body]
Touch 2: [subject] → [body]
Touch 3: [subject] → [body]

CALL SCRIPT:
[MEDDIC-structured script]

═══════════════════════════════════════════
APPROVE? (y/edit/skip)
═══════════════════════════════════════════

Stage 3: Approve

• Present draft to Tim using AskUserQuestion tool
• Options: "Approve & Load", "Edit first", "Skip this prospect"
• If edit: Tim provides changes, re-draft, re-present

Stage 4: Load to Apollo Sequence

On approval:

1. Create contacts in Apollo via apollo_contacts_create:

- first_name, last_name, email, organization_name, title

1. Find target sequence via apollo_emailer_campaigns_search:

- Search by sequence name Tim specifies (or default active sequence)

1. Add to sequence via apollo_emailer_campaigns_add_contact_ids:

- Pass contact IDs + sequence ID - Confirm enrollment count

1. Optional: Create Gmail draft via gmail_create_draft:

- For high-ICP prospects (score 80+), also create a direct Gmail draft

Sibling Skills Referenced

• sales-revenue-skill — Email templates, lead scoring tiers, MEDDIC framework
• hubspot-revops-skill — Golden Rules filter logic, HubSpot query patterns
• research-skill — Competitive intelligence, firmographic research patterns
• CLAUDE.md § ATL/BTL Classification v1.0 — Approved title keyword patterns for ATL/BTL/Gray classification (2026-03-17)

Emit Outcome Sidecar

As the final step, write to ~/.claude/skill-analytics/last-outcome-prospect-research-to-cadence.json:

{"ts":"[UTC ISO8601]","skill":"prospect-research-to-cadence","version":"1.0.0","variant":"default",
 "status":"[success|partial|error]","runtime_ms":[estimated ms from start],
 "metrics":{"prospects_researched":[n],"emails_drafted":[n],"sequences_loaded":[n],"atl_count":[n],"icp_avg_score":[n]},
 "error":null,"session_id":"[YYYY-MM-DD]"}

Use status "partial" if some stages failed but results were produced. Use "error" only if no output was generated.