clickup-reference-architecture
Analyze and design a governed ClickUp integration with typed v2/v3 adapters, Workspace policy, durable events, reconciliation, telemetry, and reversible writes. Use when shaping a production ClickUp service. Trigger with "ClickUp architecture", "design ClickUp sync", or "ClickUp integration blueprint".
Allowed Tools
Provided by Plugin
clickup-pack
Production-grade Claude Code operator workflows for governed ClickUp integrations (24 skills)
Installation
This skill is included in the clickup-pack plugin:
/plugin install clickup-pack@claude-code-plugins-plus
Click to copy
Instructions
ClickUp Integration Reference Architecture
Overview
Separate provider transport, tenancy policy, business mapping, durable execution, and evidence so API evolution or partial failure does not corrupt work.
Prerequisites
- System boundaries, source-of-truth decision, data classes, latency/freshness SLOs, and ownership
- Endpoint/version and plan inventory for required ClickUp capabilities
- Failure, reconciliation, rollback, and incident requirements
Tool Discipline
Use Read, Glob, and Grep to inspect the repository, adapters, configuration names, tests, and evidence. Use WebFetch only for current official ClickUp documentation. Use Write or Edit after confirming the target file, Workspace boundary, and requested mode.
Current Contract
- Use explicit v2 and v3 clients because ClickUp exposes only selected v3 surfaces today.
- Resolve v2
team_idto the application Workspace tenant boundary before every operation. - Treat webhooks as signed change signals processed through a durable idempotent queue, not as a complete event log.
- Writes require stable source identity, reconciliation, and compensating/rollback policy.
Authentication
Use a personal token only for accountable individual/testing work or OAuth Authorization Code for a user-facing integration. Inject the token server-side through a governed secret reference, send it in Authorization, verify authorized Workspace IDs, and never print the token, OAuth client secret, or webhook secret.
Instructions
- Define source of truth and field-level ownership for every one-way or two-way mapping.
- Place auth and Workspace authorization in a policy layer above versioned transports.
- Model commands and events with durable IDs, mapping versions, and idempotency records.
- Add rate-aware queues, webhook ingress, reconciliation sweeps, and dead-letter review.
- Separate content-free telemetry/evidence from sensitive payload storage and retention.
- Exercise auth, plan, rate, webhook-gap, partial-write, schema-drift, and rollback scenarios.
Approval Boundaries
Do not permit autonomous destructive writes, cross-Workspace routing, ACL changes, or conflict resolution without explicit product-owner policy.
Output
Return component/data-flow design, version matrix, trust and tenant boundaries, failure modes, SLOs, reconciliation, approvals, and rollback.
Error Handling
| Condition | Response |
|---|---|
| Source of truth is ambiguous | Stop two-way design until ownership is decided. |
| Required endpoint is plan-gated | Record the dependency and approved alternative. |
| Webhook-only design cannot reconcile gaps | Add source reads and durable checkpoints. |
| Compensation is impossible | Narrow write scope or require manual approval. |
Examples
The example below is a redacted operator receipt; it contains no task text, member data, credential, or webhook secret.
mode=two-way; versions=v2-tasks+v3-audit; tenant-guard=required; queue=durable; reconcile=hourly; destructive=manual