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 command-first and genuinely actionable, with a clear connection lifecycle including state-based error handling. Its main weaknesses are an unnecessary marketing-style introduction, a vague entity overview section, a broken "Step 2" cross-reference, and no bridge from the action key table to the `action run <actionId>` command.
Suggestions
Delete the marketing intro paragraph, the entity tree with "Use action names and parameters as needed", and the "Agent Types" aside — none of it changes what Claude executes.
Fix the dangling references: replace "skip to **Step 2**" with the actual target section (e.g. "Searching for actions"), and number the workflow steps consistently (1, 1b, 2, ...).
Show how to resolve an action from the Popular actions table into the `<actionId>` that `membrane action run` needs (e.g. a `membrane action list` lookup by key), and reconcile the `npm install -g` vs `npx @membranehq/cli` invocations.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The CLI instructions are tight and command-first, but the body carries unnecessary padding Claude doesn't need: the marketing intro ("a cloud-based platform that helps businesses manage and automate their marketing workflows..."), a bare entity tree ending in the filler line "Use action names and parameters as needed", and an "Agent Types" aside. Mostly efficient with several sections that should be trimmed. | 3 / 5 |
Actionability | Concrete, copy-paste-ready membrane commands throughout — install, login (including headless `login complete <code>`), `connection ensure` with a real example URL, `action list --intent`, `action run --input`, and a full proxy flag table. Minor gaps keep it below 5: the Popular actions table gives keys but never shows how to obtain the `actionId` that `action run` requires, and the install uses `npm install -g` while a later command invokes `npx @membranehq/cli`. | 4 / 5 |
Workflow Clarity | The sequence (install → authenticate → connect → wait for READY → search actions → run → proxy fallback) is clear, with explicit checkpoints: `--wait` long-polling for state, a decision tree for `CLIENT_ACTION_REQUIRED` / `CONFIGURATION_ERROR` / `SETUP_FAILED`, and re-polling after user action. The gap is navigational: "skip to **Step 2**" appears twice but no Step 2 exists — the numbering scheme is broken. | 4 / 5 |
Progressive Disclosure | A single self-contained file (~145 lines) with well-organized section headers and no bundle files to reference; nothing is deeply nested or buried. Good structure overall, though the proxy flag table and connection-state troubleshooting are candidates to move into a reference file if the skill grows. | 4 / 5 |
Total | 15 / 20 Passed |