Content
88%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 strong, densely actionable body with clear sequencing, explicit validation steps, and a final checklist — the best parts are the mode/waitForSubWorkflow semantics and the silent-contract-break warnings. Its two real weaknesses are triple repetition of key rules (prose → anti-patterns table → checklist) and two heavily referenced companion files that are missing from the bundle, leaving the worked examples and naming conventions unreachable.
Suggestions
Ship SUBWORKFLOW_PATTERNS.md and NAMING_AND_DISCOVERY.md in references/ (or inline their essentials) — the body points to them for the N+1 worked example, fire-and-forget pattern, and naming convention, but no references/ directory exists, so those deferrals are dangling.
State each rule once and let the Quick reference checklist be the recap — currently ~10 rules appear in both the prose sections and the Anti-patterns table before repeating a third time in the checklist.
Trim the motivational introduction ("badly underused", "the things functions buy you everywhere", the copy-paste-drift narrative) to one sentence — it argues for the skill rather than instructing.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body assumes Claude's intelligence — it never explains what n8n or a workflow is, and focuses on non-obvious judgment calls (mode semantics, Define Below vs passthrough, the N+1 pattern). It falls short of anchor 5 because the motivational intro ("badly underused", "copies quietly drift apart", "the things functions buy you everywhere") is rhetorical padding, and key rules are stated three times (prose, the Anti-patterns table, and the Quick reference checklist), costing tokens for reinforcement. | 4 / 5 |
Actionability | Guidance is concrete and executable throughout: copy-paste trigger JSON with typed "workflowInputs.values", exact MCP calls ("n8n_list_workflows()", "n8n_test_workflow({workflowId, method: \"prepare\"})"), exact expression syntax ("$json.list_of_ids", "$('When Executed by Another Workflow').first().json.<field>"), a specific error contract ("{ ok: false, error: \"...\" }"), and a concrete zero-input fix ("'Keep Only Set', no fields"). The common cases — declaring inputs, calling with the right mode, naming, testing — are each covered with specific examples; only the caller-side node JSON is described rather than shown. | 5 / 5 |
Workflow Clarity | The process is clearly sequenced (search first → decide whether to extract → design the contract → call correctly → verify) with explicit validation checkpoints: "validate with `validate_workflow` and `n8n_get_workflow` before you're done" for the dangerous field-rename migration, test-in-isolation instructions ("method: \"prepare\"" then "method: \"pinned\""), and a final pre-shipping checklist. This matches the top anchor: clear sequence, validation steps, error-recovery guidance, and a checklist. | 5 / 5 |
Progressive Disclosure | In-body structure and signaling are excellent: a "Reference files" table with a "Read when" column, one-level-deep pointers, and a lean core that defers depth to SUBWORKFLOW_PATTERNS.md and NAMING_AND_DISCOVERY.md. However, neither referenced file exists in the bundle (no references/ directory), so navigation breaks at exactly those points — the flagship worked examples (N+1 pattern, fire-and-forget parallelization) and the naming convention are deferred to files the reader cannot open. Good structure with a real defect; not anchor 5. | 4 / 5 |
Total | 18 / 20 Passed |