Tool order:

Read the handoff JSON; run_date must be today or the run halts (stale input).
hubspot_search_contacts (Epiphan AI, by email) + apollo_contacts_search — reuse vs skip (already sequenced) vs create.
qualify_lead (Epiphan AI) by email per prospect — Gate Q final go/no-go. Phone must be non-null.
apollo_emailer_campaigns_search — resolve sequence IDs by name from sequence_map in config.json.
apollo_email_accounts_index — resolve the mailbox for tkipper@epiphan.com; none → halt.
apollo_contacts_create (new only) → apollo_emailer_campaigns_add_contact_ids (≤100 per call) → hubspot_create_contact for new ones.
Deliver: enrollment report (chat + ~/.claude/skill-state/sequence-load/<date>.md), Slack DM to Tim on scheduled runs. </quick_start>

<house_rules> Customer-facing skill. Before output, run the gates in order (details: ~/.claude/skills/epiphan-house-rules-skill/reference/gates.md): S: search_product_catalog (Epiphan AI) or Epiphan Knowledge search before any product claim. B: check_my_copy on every customer-facing string. Q: qualify_lead right after the candidate pull, before tiering/enrichment/drafts. D: outbound text becomes a Gmail draft, never a send. Identifiers (owner IDs, portal, emails, seats) come from reference/identifiers.md in that skill — do not restate them here. On any gate failure follow reference/self-healing.md (retry, degrade with a flag, alert Tim, halt) and append a line to ~/.claude/skill-runs/.jsonl. </house_rules>

<success_criteria>

Stage 1: Read the prospect-refresh Handoff

Input: ~/.claude/skill-state/prospect-refresh/last-run.json → prospects[] with name, email, title, company, domain, vertical, icp_score, tier, apollo IDs, phone, draft_id.

Validation: run_date = today (else halt: "stale handoff — did prospect-refresh run?"), email format valid, vertical ∈ {Higher Ed, Courts, Government, Corporate AV, Healthcare, Houses of Worship, K-12}. If the file is missing after one retry, there is no safe fallback: halt, Slack DM Tim, status error (stage: input). Do not scrape the HTML report.

Output: structured list (typically 30).

Stage 2: Dedupe Against HubSpot + Apollo

A — HubSpot: mcp__claude_ai_Epiphan_Ai__hubspot_search_contacts (query = email, searchBy: "email").

Exists and not in an active sequence (hs_lead_status not Sequenced/Active Sequence) → reuse the contact ID
Exists and sequenced → SKIP ("already enrolled")
Not found → create in Stage 9

B — Apollo: mcp__plugin_sales_apollo__apollo_contacts_search (q_keywords = email).

Exists with an active sequence → SKIP
Exists without → re-enroll with the existing ID
Not found → create in Stage 5

Stage 3: Gate Q — qualify_lead final go/no-go + phone

MCP Tool: mcp__claude_ai_Epiphan_Ai__qualify_lead with email per prospect, writeToHubSpot: false. This is the last gate before a live sequence, so the verdict is authoritative even though prospect-refresh already ran it (data can change in 45 minutes, and manual runs may carry older input). Customer / channel / product-only / AE-active / suppressed / junk / NEVER → not enrolled, reason logged. STALE AE LEAD → not enrolled, listed for Tim. Rules and owner IDs: ~/.claude/skills/epiphan-house-rules-skill/reference/gates.md and reference/identifiers.md.

Phone: must be non-null (E.164 or +1-XXX-XXX-XXXX). Missing → log "Phone missing", do not enroll (Apollo compliance + Tim's dial requirement).

qualify_lead unavailable after one retry → halt (never enroll a list Gate Q has not seen), Slack DM Tim, status error (stage: gate_q).

Stage 4: Map Prospects to Sequences

MCP Tool: mcp__plugin_sales_apollo__apollo_emailer_campaigns_search (per_page: 100) → id, name, status, num_contacts.

Mapping comes from sequence_map in config.json (vertical → sequence name pattern, plus default). For each vertical pick the matching active sequence with the lowest current enrollment; none → the default; default missing → halt and tell Tim. Never hardcode sequence IDs here.

Stage 5: Create Apollo Contacts (new only)

mcp__plugin_sales_apollo__apollo_contacts_create → email, first_name, last_name, title, organization_name, phone, label_names: [vertical, "BDR_Prospect", "<YYYY_Qn>"]. Keep the returned contact ID.

Stage 6: Resolve the Sending Mailbox

mcp__plugin_sales_apollo__apollo_email_accounts_index → pick the account whose email is tkipper@epiphan.com (per identifiers.md; the running rep's own mailbox when rep-interchangeable). None found → halt before any enrollment, Slack DM Tim ("Apollo mailbox not linked"), status error (stage: mailbox). Never paste an account ID literal into this file.

Stage 7: Batch Add to Sequences (approval point for manual runs)

Manual runs: show the enrollment table (Stage 10 format, status "pending") and wait for Tim's approval via AskUserQuestion before calling Apollo. Scheduled runs: proceed (standing approval for this task) and DM the report.

mcp__plugin_sales_apollo__apollo_emailer_campaigns_add_contact_ids per sequence:

id / emailer_campaign_id: sequence_id
send_email_from_email_account_id: <Stage 6>
contact_ids: [≤100]
sequence_active_in_other_campaigns: true
sequence_unverified_email: false
status: "active"

Retry once on rate limit with backoff. If a whole batch fails twice → stop further batches, Slack DM Tim, status partial or error depending on whether any batch succeeded.

Stage 8: Confirm Enrollments

apollo_contacts_search per enrolled email → contact.sequences contains the sequence ID, enrollment date = today, status active. Log per prospect: ✓ email → sequence / ✗ email → reason.

Batch threshold: enrolled / eligible < 50%, or zero enrolled → Slack DM Tim with the failure reasons and halt before Stage 9 (do not create HubSpot records for contacts that are not actually sequenced).

Stage 9: Create HubSpot Counterparts

For prospects not found in Stage 2A: mcp__claude_ai_Epiphan_Ai__hubspot_create_contact → firstname, lastname, email, phone, jobtitle, company, lifecyclestage subscriber, hs_lead_status open; then hubspot_create_contact_note with [sequence-load] <date> sequence=<name> tier=<ATL|GRAY> vertical=<v> icp=<n> apollo_id=<id>. Company: hubspot_search_companies by domain → associate; none → hubspot_create_company. HubSpot write failure after retry → keep going, list the unsynced contacts, partial (stage: hubspot_sync).

Stage 10: Enrollment Report

Prospect	Email	Phone	Vertical	Tier	Sequence	Apollo ID	HubSpot ID	Status

Summary: processed, enrolled (n, %), Gate Q rejections by reason, phone missing, already enrolled, HubSpot synced, per-sequence counts, STALE AE LEAD list. Save to ~/.claude/skill-state/sequence-load/<YYYY-MM-DD>.md; scheduled runs DM Tim the summary line + path. Gmail drafts from prospect-refresh remain drafts for Tim's review (Gate D).

<self_healing> partial = some contacts were enrolled but a stage degraded: a batch failed after retry (remaining batches stopped), HubSpot sync failed for some contacts, or a vertical had no matching sequence and fell to the default. error = handoff missing or stale, qualify_lead unavailable, no Apollo mailbox, or batch success below 50% / zero enrolled (halt before HubSpot sync). Ladder per ~/.claude/skills/epiphan-house-rules-skill/reference/self-healing.md: retry once → degrade with a flag → Slack DM Tim → halt with state in ~/.claude/skill-state/sequence-load/. One JSON line per run to ~/.claude/skill-runs/sequence-load-skill.jsonl. </self_healing>

Emit Outcome Sidecar

As the final step, write to ~/.claude/skill-analytics/last-outcome-sequence-load-skill.json:

{"ts":"[UTC ISO8601]","skill":"sequence-load-skill","version":"<frontmatter version>","variant":"default",
 "status":"[success|partial|error]","runtime_ms":[estimated ms from start],
 "metrics":{"prospects_in":[n],"contacts_loaded":[n],"sequences_updated":[n],"duplicates_skipped":[n],
 "rejected_by_qualify":[n],"stale_ae_surfaced":[n],"phone_validation_failures":[n],"hubspot_created":[n],"batch_success_pct":[n]},
 "error":null,"session_id":"[YYYY-MM-DD]"}

Use "partial" if some stages failed but enrollments happened; error must name the stage. Use "error" only if nothing was enrolled.

Changelog

v1.0.1 (2026-10-04): Reads prospect-refresh JSON handoff (no HTML scraping); qualify_lead final gate replaces inline Golden Rules / NEVER ATL / suppression; sequence_map + schedule declared in config.json; mailbox literal removed, tkipper@epiphan.com; manual runs confirm-first; batch threshold alert + halt; self-healing + run log; frontmatter name = directory.
v1.0 (2026-03-19): Initial release.

sequence-load-skill

Sequence Load Skill