Content
81%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.
The body is a high-quality, deeply actionable guide with explicit verification steps and excellent in-body organization, and its domain knowledge is genuinely non-obvious. Its one serious defect is that the entire reference layer it is built around — four files it repeatedly defers to — is missing from the bundle, so the progressive-disclosure architecture fails at runtime; conciseness could also improve by consolidating repeated traps.
Suggestions
Ship the four referenced files (NODE_ERROR_OUTPUTS.md, API_WORKFLOWS.md, RESPONSE_SHAPES.md, ERROR_WORKFLOWS.md) in a references/ directory — or, if they are not available, inline the essential deferred content (Switch JSON, per-field validator schema, correlation-ID/retry_after conventions, full failure-mode catalog) so no instruction points to a nonexistent file.
Consolidate the repeated traps: the response-code-defaults-to-200 trap and the half-wired error-output trap each appear in the narrative, the anti-patterns table, and the checklist — state each once in depth and let the checklist reference it in a line.
Move the time-sensitive version pin ("Verified on n8n 2.38.5") into a dedicated compatibility note with re-verification guidance, or drop the pin from the main narrative, to keep the core sections date-stable.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense with n8n-specific operational knowledge Claude does not already know (onError semantics, MCP limitations, failure-mode table) with little conceptual padding. Minor over-explanation remains: the response-code-defaults-to-200 trap and the half-wired error-output trap each repeat across 3-4 sections (narrative, anti-patterns table, checklist), and the time-sensitive version pin "Verified on n8n 2.38.5" sits outside any old-patterns section. Not 5: repetition and re-editorializing mean not every token earns its place; not 3: there is no generic explanatory filler. | 4 / 5 |
Actionability | Fully executable guidance throughout: copy-paste-ready updateNode/addConnection fragments, exact retry values (retryOnFail: true, maxTries: 3, waitBetweenTries: 5000), a cause-to-status-code table with error codes, a Switch routing diagram with fallbackOutput: extra, the error-workflow trigger payload, and an explicit verification procedure via n8n_get_workflow. Common cases (wire error output, configure retry, classify and respond, build error workflow) are each covered by concrete examples. | 5 / 5 |
Workflow Clarity | Multi-step processes are clearly sequenced with explicit validation checkpoints: the two-step error-output setup (set onError, wire sourceIndex: 1) is followed by a 'Then verify' step that names exactly what to confirm because validate_workflow misses the trap; retry is ordered before error branches; the cause-to-status mapping separates 4xx-before-work from 5xx-from-error-outputs; and per-workflow-type checklists close it out. Feedback loops (test-run, read node statuses, fix and re-verify) are present. | 5 / 5 |
Progressive Disclosure | In-body structure and signaling are exemplary — a 'Reference files' table with read-when guidance, one-level-deep references, clear section flow — but the actual bundle contains only SKILL.md: no references/, scripts/, or assets/ directories exist, so all four referenced files (NODE_ERROR_OUTPUTS.md, API_WORKFLOWS.md, RESPONSE_SHAPES.md, ERROR_WORKFLOWS.md) are dead ends. Per the rubric's guideline to score against the actual bundle structure, the deferral of key content (Switch JSON, per-field validator, correlation-ID conventions, full failure-mode catalog) to nonexistent files makes navigation broken in practice, matching the anchor-2 level of a defective reference structure rather than the intact, clearly signaled structure of 3+. | 2 / 5 |
Total | 16 / 20 Passed |