n8n Subworkflow Architect
Turn reusable n8n logic into explicit, testable sub-workflows that humans and AI agents can discover and call safely.
The material distinguishes stateless and stateful sub-workflows, requires side effects to be declared, and mentions validation, structured errors, and completion tracking. It does not require user confirmation for external writes, least-privilege permissions, sensitive-data handling, credential isolation, rollback, or explicit data-flow review; publisher identity is also unverified, so points are deducted.
The structure, examples, and anti-patterns are mostly self-consistent, with guidance for expected failures and silent unrecognized inputs. However, n8n/MCP behavior is not verified within the supplied files, no executable test suite or reproducible key-path evidence is provided, and the activation guidance conflicts on the roughly 10-node versus over-5-node threshold; static scoring is therefore limited.
The audience, activation scenarios, boundaries for Define Below versus passthrough, and choices between all/each and synchronous/asynchronous execution are clearly described. Complete non-fit ranges and Chinese-language guidance are absent, and mainland-China network reachability is not demonstrated, so points are deducted.
The documentation is layered and cross-referenced, with examples, anti-patterns, checklists, version 1.0.0, compatibility notes, and an MIT license. Maintenance ownership, changelog practice, and update paths are incomplete, and the README's skill count conflicts with the supplied repository scope, so full marks are not justified.
It offers directly actionable guidance for decomposition, contracts, naming, invocation modes, and parallelization, with a clear core benefit. Yet the files contain no executed outputs or representative-result verification, and important behavior depends on an external n8n-mcp environment; static review cannot confirm complete correctness, so the score remains at or below 7.
Concrete MCP call examples, configuration snippets, decision rules, and checklists provide some auditability. The supplied files contain no key-path test suite or real CI execution evidence, so the evidence is largely author documentation and only limited verifiability is supported.
- Do not treat the documented n8n-mcp capabilities, concurrency semantics, or compatibility claims as execution-verified; test key paths in an isolated instance before deployment.
- For database, Data Table, Slack/SMTP, or other external writes, add user confirmation, scoped permissions, sensitive-data controls, idempotency, and recovery procedures.
- The publisher is not verified by the FollowSkills enterprise registry; identity should be treated as unknown.
- Activation thresholds are inconsistent, and explicit Chinese-language and mainland-China network guidance is missing.
What does this skill do, and when should you use it?
This is the standalone sub-workflow skill from the n8n-skills repository, focused on reuse, workflow modularity, and input/output contracts. It tells builders to search existing workflows before creating new logic and to use typed Define Below inputs for normal structured calls. It explains all versus each execution, blocking versus fire-and-forget calls, stateless versus deliberate stateful designs, and splitting divergent input shapes. It fits teams building complex or repeated n8n workflows through n8n-mcp, but it is not a general node-configuration, deployment, or whole-workflow design skill.
Guides the assistant to search existing workflows with n8n_list_workflows and n8n_get_workflow; configure Execute Workflow Trigger nodes with Define Below and typed workflowInputs.values; define stable input/output contracts and an explicit final Return node; choose mode all or each and decide whether waitForSubWorkflow should block; name and describe workflows for discovery; return expected failures as { ok: false, error }; validate with validate_workflow, test in isolation with n8n_test_workflow, inspect runs with n8n_executions, and use Data Tables when state tracking is needed.
- An n8n developer repeats parsing, authentication, retry, formatting, or ID-generation logic across workflows and wants one shared implementation.
- A team has a workflow with roughly 10 or more nodes and wants to isolate independently testable sections.
- A builder wants to expose a reusable multi-step process as an AI-agent tool with structured parameters.
- A workflow must support genuinely different contracts such as JSON versus binary or synchronous versus asynchronous inputs.
- A caller needs per-item execution, background dispatch, or completion tracking and must choose mode, waitForSubWorkflow, and a Data Table pattern deliberately.
What are this skill's strengths and limitations?
- Makes search-before-build a core operating rule.
- Provides concrete guidance for typed inputs, natural output shapes, and contract changes.
- Covers all/each execution, blocking/background calls, and N+1 input-shape splitting.
- Treats both pure reusable logic and explicitly documented stateful operations as valid designs.
- Names specific tools for validation, isolated testing, execution inspection, and state tracking.
- Requires the n8n-mcp MCP server and cannot perform these operations independently of it.
- Binary inputs generally require passthrough and do not wire cleanly as agent-tool inputs.
- Zero-input operations also require passthrough and additional cleanup/documentation discipline.
- The source does not document a standalone test command or standalone package for this skill.
- Its scope is sub-workflow design; it does not replace general node configuration, deployment, or whole-workflow pattern guidance.
How do you install this skill?
Install and configure the n8n-mcp MCP server first. For Claude Code, the repository recommends installing the collection with: /plugin install czlonkowski/n8n-skills. For manual installation, run git clone https://github.com/czlonkowski/n8n-skills.git, then copy n8n-skills/skills/* to ~/.claude/skills/. The source does not provide a separate installation command for this individual skill.
How do you use this skill?
In Claude Code or Claude.ai with n8n-mcp configured, try a prompt such as: “Extract this repeated n8n logic into a reusable sub-workflow with typed Define Below inputs.” Mentioning “sub-workflows,” “Execute Workflow,” “mode each vs all,” or “waitForSubWorkflow” also matches the skill description.
How does this skill compare with similar options?
Compared with n8n’s Custom Code Tool, a sub-workflow is positioned for reusable, multi-step logic shared across agents or workflows, while the Custom Code Tool is the inline agent-tool option.