Content
75%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 strong, command-driven integration guide: install, authenticate, connect, poll for readiness, discover actions, run them, and fall back to a proxy — all with executable commands and explicit connection-state checkpoints. The main weaknesses are a padded 43-item entity list that wastes context, a broken "Step 2" cross-reference, and the absence of post-run verification guidance for destructive operations.
Suggestions
Cut the 43-item entity bullet list (or compress it to one line listing example tables per object family); it conveys no actionable information and Claude can discover the full entity set via `membrane action list`.
Fix the workflow navigation: label the step after connection setup as "Step 2: Find and run actions" so the "skip to Step 2" references resolve, and use `membrane` (or `npx`) consistently for the polling command.
Add a brief validation checkpoint for destructive operations, e.g. "Before running delete-records, run get-record to confirm the target IDs, and verify the count in the response before proceeding."
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The 43-item entity bullet list ("Contacts", "Deals", ... "Expense Categories") carries no explanatory value and is pure padding, and lines like "Use action names and parameters as needed" add nothing. The rest (CLI install, auth, connection, actions, proxy, best practices) is efficient and command-focused, so this sits at 'mostly efficient but includes some unnecessary content' rather than the severely padded anchor 2. | 3 / 5 |
Actionability | Fully executable, copy-paste-ready commands throughout: `npm install -g @membranehq/cli@latest`, `membrane connection ensure "https://companyhub.com" --json`, `membrane action run <actionId> --connectionId=CONNECTION_ID --input '{"key": "value"}'`, plus a proxy flags table and a popular-actions table with exact action keys. The common cases are concretely covered. | 5 / 5 |
Workflow Clarity | The sequence (install → login → connection ensure → poll state → search actions → run → proxy fallback) is clear with explicit state checkpoints (READY / BUILDING / CLIENT_ACTION_REQUIRED / CONFIGURATION_ERROR) and re-poll guidance. Not 5: "skip to **Step 2**" references a step that is never labeled, `npx @membranehq/cli` and `membrane` are used inconsistently for the same command, and there is no post-run verification guidance despite covering destructive actions like Delete Records. | 4 / 5 |
Progressive Disclosure | A single-file skill (no references/, scripts/, or assets/ exist) with well-organized headers, a clearly signaled popular-actions table, and appropriately placed command examples. Minor gap: the 43-entity list is inlined bulk content that adds navigation noise rather than value, keeping it below the clean-overview anchor at 5. | 4 / 5 |
Total | 16 / 20 Passed |