011matthias

make-pack

Consolidated Make.com skill pack. Use when building, editing, testing, or debugging Make.com scenarios. Replaces make-mcp-tools-expert, make-scenario-patterns, webhook-inspector, and blueprint-reconciler. Load modules individually per task.

011matthias 0 Updated 3mo ago
GitHub

Install

npx skillscat add 011matthias/agentic-ops1-01/make-pack

Install via the SkillsCat registry.

SKILL.md

Make.com Pack

Unified reference for building Make.com scenarios. Consolidates: make-mcp-tools-expert, make-scenario-patterns, webhook-inspector, blueprint-reconciler.


Build Procedure

  1. Detect — Confirm Make.com orchestrator (infrastructure.yaml has type: make)
  2. Read spec — Extract flow, systems, edge cases, acceptance criteria
    2.5. Estimate ops — For each scenario, estimate monthly operations before building. Scheduled: (2,592,000 / interval_seconds) × modules_in_flow. Webhook: events_per_month × modules_in_flow. Compare total against client's Make.com plan — if projected > 80% of limit, flag before building. Load OPERATIONS-ANALYZER Section A for full estimation procedure.
  3. List connections — Verify auth connections exist for required services
  4. Generate blueprint → Load BLUEPRINT module for JSON format
  5. Deploy — Try MCP scenarios_update first; fall back to uv run tools/make-api.py update|deploy or UI import if MCP fails.
  6. Post-deploy binding check — Load POST-DEPLOYMENT-VERIFICATION after API deployment. Alert user to rebind data store/connection modules in UI.
  7. Verify — Test with Run Once, check execution inspector
  8. Activate — Enable scheduling or webhook listening

Critical Rules (Always Apply)

  • Check API impossibilities FIRST — Before attempting any API approach, load API-IMPOSSIBILITIES. Gmail/OAuth scenarios cannot be made functional via API alone. State UI requirements in your FIRST response — do not iterate.
  • MCP blueprint deployment may work — scenarios_update with blueprint param succeeded for A3 on 2026-03-14 after previously failing with 500. Test each deployment; fall back to REST API (tools/make-api.py) or UI import if it fails.
  • Module name casing matters — datastore:AddRecord (capital A), not datastore:addRecord
  • Connection IDs are instance-specific — cannot copy between orgs
  • Blueprint as JSON string for REST API (MCP may auto-stringify)
  • Key terminology (vs n8n): Scenario=Workflow, Module=Node, Route=Connection, Connection=Credential, Router=IF/Switch

Module Index

Load ONE module at a time based on your current task.

Procedure Modules

When Module Source
Generating blueprint JSON BLUEPRINT-FORMAT make-mcp-tools-expert
Setting up a new project PROJECT-SETUP make-mcp-tools-expert
Testing webhooks WEBHOOK-TESTING make-mcp-tools-expert
Diagnosing scenario failures AUTONOMOUS-DIAGNOSTICS make-mcp-tools-expert
Verifying execution outcomes POST-EXECUTION-VERIFICATION make-mcp-tools-expert
After API blueprint deployment POST-DEPLOYMENT-VERIFICATION make-mcp-tools-expert
Iterating on fixes ITERATION-CYCLE make-mcp-tools-expert
Adding/modifying data store fields SCHEMA-EVOLUTION make-mcp-tools-expert
Discovering webhook payloads WEBHOOK-PAYLOAD-INSPECTOR make-mcp-tools-expert
Pre-handover checklist PRE-CLIENT-REVIEW make-mcp-tools-expert
Cross-validating blueprint vs data stores DATA-STORE-RECONCILER blueprint-reconciler
Validating Sheets column refs SHEETS-COLUMN-RECONCILER blueprint-reconciler
Checking IML {{N.field}} refs IML-REFERENCE-CHECKER blueprint-reconciler
Validating email template placeholders TEMPLATE-PLACEHOLDER-CHECKER blueprint-reconciler
Pre-handover blueprint format check HANDOVER-FORMAT-CHECKER blueprint-reconciler
Capturing webhook payloads CAPTURE-PATTERN webhook-inspector
Analyzing captured payloads ANALYZE-PAYLOAD webhook-inspector
Estimating or analyzing operations costs OPERATIONS-ANALYZER make-mcp-tools-expert
Onboarding platform feasibility check PLATFORM-FEASIBILITY make-mcp-tools-expert
Before attempting any API approach API-IMPOSSIBILITIES make-mcp-tools-expert

Reference Modules (load ONLY for specific lookups)

When Module Source
IML expression error IML-GOTCHAS make-mcp-tools-expert
Known webhook provider formats KNOWN-PROVIDERS webhook-inspector

Scenario Architecture Patterns (from make-scenario-patterns)

Pattern Trigger Use When
Webhook Processing gateway:CustomWebhook Form submissions, CRM events, payment hooks. Set instant: true. Always send WebhookResponse.
Scheduled Sequence builtin:BasicScheduler Follow-ups, batch processing, periodic syncs. Use Sleep inside iterators.
Email + Reply Detection gmail:watchEmails Monitoring inbox, stopping follow-ups on reply. Update tracking immediately.
Conditional Branching builtin:BasicRouter Different actions by lead value/type/urgency. Routes evaluated in order; last = fallback.
Data Sync Scheduler or Watch Keeping two systems in sync. Use searchRows before addRow to prevent duplicates.

MCP Server Setup

{
  "mcpServers": {
    "make-{client}": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://<ZONE>/mcp/u/<TOKEN>/sse"]
    }
  }
}

Known API Limitations

  • No module-level I/O for successful executions — executions_get_detail returns status + counts ONLY, not module output values. Workaround: add a temporary datastore:AddRecord module after the step you want to inspect; read it back via data-store-records_list. This is the data store write-back pattern — the only way to surface intermediate values autonomously. Do NOT waste operations guessing via repeated scenario runs.
  • No sheet cell reading via MCP RPCs — use direct Google Sheets API
  • validate_blueprint_schema validates API format only — not UI import format. Run HANDOVER-FORMAT-CHECKER before handover.
  • Blueprint deployment tool: uv run tools/make-api.py {update|deploy|get|list|ds-upsert|scenario-run} — full REST API wrapper that bypasses MCP 500 errors on blueprint params.
  • Email template import: uv run tools/docx-to-datastore.py {extract|validate|upload} — converts .docx email templates to data store records. Validates ##placeholder## patterns.

Categories