Deepline Enrich

/deepline-enrich — Waterfall enrichment + email validation

Enrich prospect lists across 15+ data providers (Apollo, Hunter, PDL, Prospeo, Icypeas, ZeroBounce, BetterContact, etc.). Deepline runs a waterfall — first provider to return a valid result wins. Typically finds 20-40% more emails than any single provider.

Imported via: /steal analysis of https://code.deepline.com (2026-03-31). Phase F upgrade: Play patterns + credit gate (2026-04-04). Apify waterfall slot added 2026-05-01.

Doctrine inherited (Step 7 — 0626 rollout)

Output complies with:

• output-tenets.md — the seven tenets
• output-simplicity.md — length caps, three-layer source placement, robot-tells ban
• Step 6 calibration: see [[feedback_execution_doctrine_refinements_step6]]

Refinements applied to this skill:

Note: this skill is internal-reference only; R2/R5/R6/R7/R8 do not apply (no customer-facing surface).

Prerequisites

• Deepline CLI: curl -s "https://code.deepline.com/api/v2/cli/install" | bash
• Python 3.10+ required (macOS ships 3.9 — install via brew install python@3.13)
• Node.js 20+ required (already met)
• BYOK keys configured: Run deepline quickstart to connect existing provider API keys

Credit gate — voice-locked

ALWAYS run before any enrichment over 5 contacts. This is the load-bearing safety mechanism preventing accidental spend.

Pre-execution checklist

1. ESTIMATE cost: deepline enrich --input file.csv --dry-run
2. CHECK balance: deepline billing --balance
3. CONFIRM with user: "This will cost ~$X for N contacts. Proceed?"
4. SET spending cap: deepline billing --set-monthly-limit [amount]

Gate rules

Cost reference

Key: Deepline charges only on successful matches. Failed lookups don't consume credits.

Pilot mode — ALWAYS test first

Before any batch, test on 2 rows:

deepline enrich --input prospects.csv --output test.csv \
  --with email=person_search_to_email_waterfall:{"title":"{{title}}","company":"{{company}}"} \
  --rows 0:1

Review output. If results look right, run the full batch.

Apollo-side credit gate also applies per .claude/rules/apollo-credits.md. Apify-slot calls flow through .claude/rules/apify-credits.md.

When to use which enrichment tool — voice-locked

Rule of thumb: Deepline = MAXIMUM EMAIL COVERAGE across a list. Apollo MCP = quick lookups. Clay MCP = company-level enrichment + AI research.

Handoff pattern: /clay-search (discover) → /deepline-enrich (enrich + validate) → /outreach-emails (write sequences)

Inputs

Process

Standard waterfall flow: dry-run → pilot (2 rows) → full batch → email validation → filter to valid. Multi-stakeholder play (~30% win rate vs ~5% single-threaded) and quarterly CRM hygiene play (30% job-change rate per year) follow the same gate discipline. Full commands + workflows + provider configuration in the premium reference.

Apify waterfall slot for LinkedIn-URL-only inputs (dev_fusion/Linkedin-Profile-Scraper $10/1k or apimaestro/.../no-cookies $5/1k) also documented in the premium reference.

Plays

22 pre-built enrichment plays organized by input type: Email Finding (6), People Finding (6), Validation & Scoring (3), CRM Hygiene & Job Changes (4), Outbound Integration (3). Decision tree + per-play input/output/cost in the premium reference.

Quality

Pre-execution checks cover cost discipline (dry-run + cap + user confirmation), pilot discipline (2-row test before full batch), input quality (column normalization, dedupe), and output quality (validation step ran, valid/catch-all/invalid segmented). Common-mistakes table (no dry-run, skipped validation, using Deepline for discovery) + worked example (250-row $38 estimate) + anti-examples + quality gate (≥75% find rate, ≥80% valid, ≤$0.20/validated, ≤2% bounce) in the premium reference.

Hit-rate gate (proceed/stop) — added 2026-05-05 via /steal from Andytoizer/agentoperator-outbound-engine:

• valid + catch_all together = hit. Most deliverable catch-all domains accept real emails even when the provider can't verify the exact mailbox. Treating catch_all as a miss undercounts deliverable coverage by ~15-25% in B2B SaaS lists.
• 80% combined hit rate is the proceed/stop gate. If (valid + catch_all) / total < 0.80, stop before proceeding to research/drafting. Below 80% means upstream is broken: domains malformed, LinkedIn URLs stale, or waterfall providers misconfigured. Don't waste research and drafting time on contacts you can't email.
• If you need stricter (valid only), override explicitly in the play config — but document why in goals/MMYY-NN-cycle.md so the next operator knows the gate was tightened.

Handoff Patterns

Integration with the engagement workflow

Assigned to the Researcher role-agent. Slots into:

• Sales pipeline: "discovery prep" — auto-enrich prospects before discovery calls
• New-client onboarding: No direct role (research phase uses company-context, not contact enrichment)
• Outbound: Core enrichment step after target list build

Final ship gate

Run /premortem --output before ship. See /premortem skill for the 5 execution domains (will-it-resonate / will-it-convert / will-it-stay-on-brand / will-stakeholder-push-back / will-it-degrade-over-time) and output template.

Trivial-case escape: ## Premortem\nNo failure modes — trivial change satisfies the contract for genuinely trivial outputs.