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, opinionated router body: it routes via two lookup tables, hard-codes validation into the protocol, and gives genuinely executable guidance throughout with almost no token waste. The main gaps are the length and drift-prone version numbers concentrated in the inline tool inventory — content that partially belongs behind the specialist-skill boundary it otherwise respects.
Suggestions
Move per-tool parameter detail and version gates (e.g. `detail`/`mode` values, canvas-group ops, folder/Data Table caveats) into `n8n-mcp-tools-expert` and keep only tool names and one-line purposes here, cutting the tools section roughly in half.
Consolidate scattered version requirements (2.28+, 2.97+, ≥2.19, 2.34+, ≥2.30/2.32) into a single version-availability note near the drift policy so they are easier to maintain and don't age silently.
Trim rhetorical asides (e.g. "it's all markdown — edit the skill", the plugin-hooked caveat repetition) to tighten the closing sections.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense and expert-aimed — decision tables ("| Job | Use |"), a red-flags routing table, and terse rules like "A Set that only re-maps fields for the next node is the antipattern" with no explanations of concepts Claude already knows. It sits between anchors 4 and 5: mostly every token earns its place, but the ~60-line tool inventory, scattered version gates ("n8n 2.28+", "n8n-mcp 2.97+", "≥ 2.19", "2.34+", "≥ 2.30", "≥ 2.32") that will drift, and closing asides ("it's all markdown — edit the skill") are trimmable over-explanation. Noticeably above the midpoint, so 4 rather than 5. | 4 / 5 |
Actionability | Guidance is fully concrete and executable for an instruction-only router: exact tool names with argument shapes ("`{id, operations:[…]}`", "`{topic, depth:"full"}`"), a job→node decision table ("| Paging an API (no service node) | HTTP Request's pagination option |"), a thought→skill routing table, and a named Code-justification convention ("`Why Code: <native node considered> — <reason>`"). Matches the score-5 anchor for specific examples covering the common cases; absence of code is not penalized since this is an instruction/routing skill and the guidance is actionable. | 5 / 5 |
Workflow Clarity | "The protocol, in order" gives a clear 5-step sequence with explicit validation checkpoints ("`get_node` before configuring any node", "**`validate_workflow` before activating** and **`n8n_get_workflow` after** to check `connections`"), and Non-negotiable #2 embeds a feedback loop (validate → verify → inspect output values because "JS errors inside `{{ }}` resolve silently to `null`"). "When in doubt" adds error-recovery paths. Matches the score-5 anchor for clear sequence, explicit validation, and feedback loops for error recovery. | 5 / 5 |
Progressive Disclosure | Structure is good: a skill-index table with one-level-deep, well-signaled references ("See `n8n-mcp-tools-expert`", "See `n8n-expression-syntax`") and easy navigation via the red-flags table. It falls short of a 5 because the "n8n-mcp tools — working knowledge from turn one" section inlines ~60 lines of working reference (parameter shapes, version gates, per-tool caveats) that overlaps the delegated `n8n-mcp-tools-expert` territory — content that is a candidate for the separate file per the score-4 anchor ('most content is appropriately placed; references mostly clear; minor organization gaps'). Note: no references/, scripts/, or assets/ directories exist in this bundle, and the referenced sibling skills are pack-level, so references could be verified only as names, not files. | 4 / 5 |
Total | 18 / 20 Passed |