Content
65%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 content is a mostly actionable CLI integration guide with executable commands and a sensible flow from setup to running actions and proxying raw requests. Its main weaknesses are incoherent step references (a missing "Step 2" heading, an unanchored "1b"), and an unresolved identity mismatch — the intro calls Arive a returns management platform while the overview lists Trip/Leg and the popular actions are Leads and Loans.
Suggestions
Fix the broken workflow references: rename the sections into a coherent numbered sequence (Step 1: connect, Step 1b: wait for ready, Step 2: search/run actions) so "skip to **Step 2**" has a real target.
Resolve the identity mismatch — the intro says Arive is "a returns management platform for e-commerce" while the Arive Overview lists Trip/Leg and Popular actions are Leads/Loans; pick one product description and align the entity list and action table with it.
Move the detailed connection clientAction state machine and the proxy request flag reference into a references/ file, keeping SKILL.md as a lean overview that links to them.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is mostly efficient: CLI commands are given as copy-paste snippets with one-line explanations ("membrane connection ensure", the action list/run examples, the proxy flag table). A few sentences pad rather than instruct ("so you can focus on the integration logic rather than auth plumbing", "This will burn less tokens and make communication more secure"), which keeps it at anchor 4 (efficient, minor instances that could be trimmed) rather than 5. | 4 / 5 |
Actionability | Concrete, executable commands appear for every stage — install ("npm install -g @membranehq/cli@latest"), login, connection ensure/poll, action search/run with --input JSON, and proxy requests — plus a popular-actions table with keys. It is not a 5 because placeholders like <actionId> and CONNECTION_ID are never shown being resolved end-to-end and one poll command switches from "membrane" to "npx @membranehq/cli" without explanation. | 4 / 5 |
Workflow Clarity | A rough sequence exists (install → login → connect → poll for READY → search actions → run), and the connection-state section does describe outcomes (READY / CLIENT_ACTION_REQUIRED / CONFIGURATION_ERROR), but the step references are incoherent: "skip to **Step 2**" points at a heading that does not exist, and "#### 1b" appears with no numbered Step 1 or 1a. This matches anchor 3 (sequence present but checkpoints and structure are implicit or broken). | 3 / 5 |
Progressive Disclosure | The body is well-sectioned with clear headers, but it is a single ~150-line file with no bundle/reference files, and content that would fit a reference file is inlined — notably the full connection clientAction state machine and the proxy flag reference table. That matches anchor 3 (some structure, but content that should be separate is inline) rather than 4. | 3 / 5 |
Total | 14 / 20 Passed |