Content
68%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.
A strongly actionable, command-first integration guide with an unusually good connection-state workflow (explicit polling, clientAction handling, and error states). Its main flaws are the dead 20-row actions table wasting tokens, a broken "Step 2" cross-reference, and the absence of any validation around destructive operations like delete and send-campaign, which caps workflow clarity.
Suggestions
Replace the "No description" descriptions in the Popular actions table with one-line purpose summaries, or move the table to a reference file and inline only the top 5 actions.
Fix the "skip to **Step 2**" reference — no Step 2 exists; number the workflow steps (install → login → connect → search → run) or use section names.
Add a verification checkpoint before destructive or irreversible actions, e.g. confirm with the user before running delete-* actions or send-campaign, and check the output field for success after running.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The CLI guidance ("Install the Membrane CLI", "membrane connection ensure", "membrane action run") is efficient and commands-first, but there is real waste: a 20-row "Popular actions" table where every Description cell reads "No description", plus an intro paragraph ("Marketers and business owners use Automizy to create and automate email campaigns...") that pads rather than instructs. Not anchor 4 because the dead table alone is ~20 wasted lines. | 3 / 5 |
Actionability | Every step ships a copy-paste-ready command: "npm install -g @membranehq/cli@latest", "membrane login --tenant --clientName=<agentType>", "membrane connection ensure \"https://Automizy.com\" --json", "npx @membranehq/cli connection get <id> --wait --json", "membrane action run <actionId> --connectionId=... --input '{\"key\": \"value\"}'", plus a complete proxy flag table. The common cases (install, auth, headless auth, connect, wait, search, run, proxy) are each concretely covered. | 5 / 5 |
Workflow Clarity | The connection flow has genuine validation checkpoints and a feedback loop (poll with --wait until state leaves BUILDING, act on clientAction, re-poll to READY, check error on CONFIGURATION_ERROR/SETUP_FAILED) — anchor 4 quality. However the skill documents destructive and outward-facing operations (delete-contact, delete-campaign, send-campaign) with no verification or confirmation steps anywhere, which caps workflow clarity at 3; the broken "skip to **Step 2**" reference (no Step 2 exists) further muddies the sequence. | 3 / 5 |
Progressive Disclosure | No bundle files exist, and the single body is well-sectioned (Install, Authentication, Connecting, Searching, Running, Proxy, Best practices) with no nested references. Not anchor 5 because the "Popular actions" table and proxy reference material are inline bulk that belongs in a separate reference file, and the "Agent Types" line is an unstructured aside. | 4 / 5 |
Total | 15 / 20 Passed |