Content
48%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 CLI usage guidance is genuinely actionable and the connection workflow is well-sequenced with state checkpoints, but these strengths are buried under an enormous padded bullet list of mostly fictitious 'entities' that wastes the context window. The skill needs radical trimming of the Overview section.
Suggestions
Delete the ~480-line entity bullet list (Lead through Better World); most are not real Zendesk Sell objects and it is pure padding that violates token efficiency.
Move any genuine entity reference content into a references/ file and link to it one level deep, or replace the list with a short prose summary of the core objects (Lead, Contact, Deal, Task, Note).
Add a brief result-verification note after 'membrane action run' so the workflow has an explicit feedback checkpoint on the action-execution side.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Roughly 480 of ~640 lines are a padded bullet list of 'entities' running from Lead/Contact through Love, Happiness, Meaning, Purpose, Compassion, and Better World — heavily padded junk, most of which are not real Zendesk Sell objects, matching the 'severely verbose; heavily padded' anchor; the lean CLI section does not rescue it. | 1 / 5 |
Actionability | The working section gives concrete, executable commands (npm install -g @membranehq/cli@latest, membrane login, membrane connection ensure, membrane action list/run, membrane request) with flags and JSON examples, fitting 'mostly executable guidance; concrete commands with minor gaps'; not 5 because parameter construction for action run is only sketchily shown. | 4 / 5 |
Workflow Clarity | The connection flow is a clear sequenced process (install -> login -> connection ensure -> poll --wait -> handle CLIENT_ACTION_REQUIRED -> search -> run) with explicit state checkpoints (READY/BUILDING/CONFIGURATION_ERROR) and a poll-again feedback loop, matching 'clear sequence with most checkpoints present'; held at 4 because result verification after running actions is not addressed. | 4 / 5 |
Progressive Disclosure | Section headers exist (Working with, Popular actions, Best practices) but the Overview is a ~480-line inlined entity list that belongs in a separate reference file or should be removed, and no bundle files exist to offload it, matching 'some structure but content that should be separate is inline'. | 3 / 5 |
Total | 12 / 20 Passed |