Salesloft Cadence Outcome Reconciliation

Overview

This skill synchronizes cadence membership state and its documented engagement counts without relying on obsolete activity paths. It preserves mutable-record semantics and a restartable cursor.

Prerequisites

• A named team and destination dataset
• cadences:read plus any endpoint-specific read scope
• A durable microsecond-precision updated_at cursor
• A documented merge key and data-retention policy

Tool Discipline

Use Read, Glob, and Grep to inspect schemas, sync checkpoints, and consumers. Use WebFetch only for current Salesloft endpoint and polling documentation. Use Write or Edit after confirming the destination contract.

Current Contract

• GET /v2/cadence_memberships supports updated_at filters, sort_by=updated_at, direction, page, and per_page up to 100.
• A membership is mutable and reused when a person later returns to the same cadence.
• Current state includes processing, staged, active, scheduled, pending, completed, removed, reassigned, and archived outcomes.
• Documented counts include views, clicks, replies, calls, sent emails, and bounces.

Authentication

Use the explicit team credential with the minimum read scopes. Never mix cursors or destination partitions between teams.

Instructions

1. Define the destination key, accepted fields, retention window, and team partition.
2. Start from a durable cursor with a deliberate overlap window for equal timestamps.
3. Fetch memberships ordered by updated_at ascending with a bounded page size.
4. Upsert by stable membership ID and treat state and counts as replaceable snapshots.
5. Advance the cursor only after the destination transaction succeeds.
6. Deduplicate overlap records and compare source, processed, and committed counts.
7. Reconcile periodic samples against Salesloft and alert on gaps or backward cursors.

Approval Boundaries

Keep this workflow read-only in Salesloft. Do not infer immutable cadence episodes, individual email events, or user performance conclusions from aggregate membership counts alone.

Output

Return team partition, cursor interval, pages and endpoint cost, source/upsert/deduplicated counts, reconciliation result, and next durable cursor.

Error Handling

Examples

The example below shows the minimum redacted evidence expected from a successful invocation of this operator workflow.

team=team-42; fetched=100; upserted=98; overlap=2; cursor=committed

Resources

• Skill-specific official documentation
• List cadence memberships
• Efficient cursor poller