Content
77%Weight 40%Scale 1-5Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
A high-signal, highly actionable design guide with excellent concrete node/connection/config detail and a strong validation-plus-checklist workflow. Its two weaknesses are length that could be trimmed into reference files, and a progressive-disclosure structure that promises nine reference files which are not actually bundled.
Suggestions
Ship the nine referenced reference files (TOOLS.md, SUBWORKFLOW_AS_TOOL.md, SYSTEM_PROMPT.md, STRUCTURED_OUTPUT.md, MEMORY.md, HUMAN_REVIEW.md, CHAT_AGENT_PATTERNS.md, RAG.md, EXAMPLES.md) under a references/ directory, and convert the bolded name references into real links (e.g. [TOOLS.md](references/TOOLS.md)) so progressive disclosure is functional.
Move explanatory scaffolding — the "Why split it" rationale, the intro framing paragraph, and the "What's NOT available via the community MCP" table — into the relevant reference files or a sibling skill to reduce SKILL.md length toward a lean overview.
Trim prose that restates adjacent tables (e.g. the anti-patterns restated in narrative form) so each guideline appears once.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Mostly efficient and dense with genuine n8n-specifics (node names, connection types, defaults), but ~280 lines with trimmable framing ("the most common over-build"), a non-core "What's NOT available via the community MCP" table, prose restating table contents, and explanatory paragraphs like "Why split it" that could move to reference files. | 3 / 5 |
Actionability | Highly concrete and executable throughout: real node types (.agent, .outputParserStructured, .toolWorkflow), real ai_* connection types, a copy-paste-ready JSON sub-node wiring snippet, exact $fromAI syntax, and concrete config values (maxIterations 15/50-200, contextWindowLength 5→50, autoFix: true, schemaType: 'manual'). | 5 / 5 |
Workflow Clarity | Pairs risky operations with explicit validation/feedback loops — validate with validate_workflow then verify with n8n_get_workflow (sub-nodes on ai_*, not main), the autoFix parse-failure retry loop, and a complete pre-shipping Quick reference checklist with checkboxes mapping to anchor 5's checklists-for-complex-processes. | 5 / 5 |
Progressive Disclosure | Structure and the "Reference files" read-when table are well designed and one-level-deep, but the 9 referenced files (TOOLS.md, SYSTEM_PROMPT.md, MEMORY.md, etc.) do not exist — there is no references/ directory — and references are bare bolded names rather than resolvable file paths, so the promised split content is absent and navigation is not functional. | 3 / 5 |
Total | 16 / 20 Passed |