Content
86%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 lean, well-structured instruction-only skill that respects the token budget and gives concrete, executable guidance for the happy path and the main failure modes. The remaining gaps are the unspecified per-tool config locations in the fallback and the absence of an OAuth-failure recovery step.
Suggestions
In the section 1 fallback, name the per-tool MCP config locations (e.g., the Claude config file, Cursor's mcp.json path) so the manual route is fully executable.
Add a brief recovery step in section 2 for OAuth failure (e.g., what to check or retry before falling back to manual config).
Include a verification command for the happy path (e.g., 'claude mcp list' or the tool's equivalent) so connection success can be checked programmatically, not just in the browser flow.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Every line carries non-obvious information (the command-shape table, PowerShell 5.1 vs 7.3 quoting rules, 'Do not paste tokens. Do not scrape the CLI keychain.') with no padding or explanations of concepts Claude already knows. | 5 / 5 |
Actionability | Guidance is mostly executable: run the command verbatim, exact Windows quoting transformation, and a fallback that names the server 'cognite'. Minor gap: the manual-config fallback omits per-tool MCP config file locations, so it is not fully copy-paste ready in the failure case. | 4 / 5 |
Workflow Clarity | A clear three-step sequence with error-recovery paths (truncated-prompt detection, command-failure fallback) and a success checkpoint ('Success is connected in the tool, not a file on disk'). Not destructive or batch, so no cap applies; OAuth-failure recovery is left unspecified. | 4 / 5 |
Progressive Disclosure | A single-purpose skill (~55 lines) with no need for external references: well-organized numbered sections, appropriately self-contained content, and one clearly signaled cross-skill pointer ('Graph reads in app code: dm-graph-traversal'). | 5 / 5 |
Total | 18 / 20 Passed |