Content
67%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 practical, command-driven guide that covers setup, authentication, connection lifecycle, action discovery and execution, and a proxy fallback — largely executable and well sequenced. Its main weaknesses are filler lines that waste tokens, the absence of a worked end-to-end action example, and a dangling "Step 2" reference with no error handling on the run step.
Suggestions
Remove the zero-value filler: the dangling "Use action names and parameters as needed." line and the bare nested entity tree (either add one-line descriptions per entity or delete it).
Add one worked end-to-end example that runs a real action key from the Popular actions table (e.g. list-registrations) with concrete input JSON and an example of the `output` field's shape.
Fix the dangling "Step 2" reference by adding explicit numbered step headings, and add brief error-handling guidance for failed action runs (checking the error field, retrying).
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dominated by real, Membrane-specific commands Claude could not know, but contains zero-value filler such as "Use action names and parameters as needed." and a bare nested entity tree with no descriptions. Mostly efficient but could be tightened — matching anchor 3 rather than 4, which requires only minor trims, and rather than 2 since no basic-concept padding is present. | 3 / 5 |
Actionability | Concrete copy-paste-ready CLI commands appear throughout — `npm install -g @membranehq/cli@latest`, `membrane connection ensure "https://www.amilia.com/" --json`, `membrane action run <actionId> --connectionId=... --input '{"key": "value"}'`, plus a proxy flag table. This matches anchor 4 (mostly executable with minor gaps) rather than 5, because no worked end-to-end example runs a real action key from the Popular actions table with realistic input and output. | 4 / 5 |
Workflow Clarity | The sequence install → login → ensure connection → poll state → handle clientAction → search actions → run → proxy fallback is clear, with explicit state checkpoints (READY / CLIENT_ACTION_REQUIRED / CONFIGURATION_ERROR) and a re-poll feedback loop. Matches anchor 4 rather than 5 because there is a dangling "Step 2" reference with no numbered heading, and no error-handling guidance after running actions. | 4 / 5 |
Progressive Disclosure | This is a single-file skill (no references/, scripts/, or assets/ exist) with well-organized headers and all content navigable in one pass — good structure, matching anchor 4. Not 5 because at ~160 lines it exceeds the simple-skill exception, and generic Membrane material (connection state machine, proxy flags) could be externalized or trimmed. | 4 / 5 |
Total | 15 / 20 Passed |