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.
The body is highly actionable with concrete, executable Membrane commands covering connection, action discovery, and proxying, but it is held back by a redundant CI/CD intro and a workflow whose destructive proxy path lacks verification plus a dangling "Step 2" reference. Progressive disclosure is solid for a single-file skill.
Suggestions
Remove or trim the introductory paragraph explaining what Semaphore/CI-CD is — Claude already knows this — and cut rhetorical padding like "so you can focus on the integration logic rather than auth plumbing".
Add a verification checkpoint after running actions and especially after proxy requests (e.g., inspect the `output`/response status, confirm the resource changed, handle non-2xx) so destructive methods are not fired blindly.
Fix the dangling "skip to Step 2" / "1b" labeling by adding an explicit numbered Step 1 / Step 2 structure for the connect-and-wait flow.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is mostly command-dense and efficient, but opens with "Semaphore is a CI/CD platform that automates software testing and deployment..." — a concept Claude already knows — and includes flourishes like "so you can focus on the integration logic rather than auth plumbing". This fits 'mostly efficient but includes some unnecessary explanation', above 2 (not heavily padded) but below 4 (the intro and rhetorical padding remain). | 3 / 5 |
Actionability | It provides fully executable, copy-paste-ready commands for every common case — `npm install -g @membranehq/cli@latest`, `membrane connection ensure`, `connection get <id> --wait --json`, `action list/run`, and `membrane request` — plus a concrete proxy flag table. Placeholders like CONNECTION_ID and <actionId> are clearly signaled, matching the 'fully executable, copy-paste ready, covers common cases' anchor. | 5 / 5 |
Workflow Clarity | There is a real sequence with state-polling feedback loops (poll until not BUILDING, re-check after clientAction, inspect `error` on CONFIGURATION_ERROR), but it is capped at 3: the proxy path supports destructive methods (DELETE/PUT) with no verification of the result, and "skip to Step 2" references a Step 2 header that does not exist. It is above 2 (sequence and some checkpoints exist) but cannot exceed 3 under the destructive-operation cap and the dangling step reference. | 3 / 5 |
Progressive Disclosure | The file is well-organized into clear sections (Overview, Working with Semaphore, Connecting, Searching, Popular actions, Proxy, Best practices) with no nested references. It is below 5 because some inline material — the detailed `clientAction` field breakdown and the proxy option table — could live in a one-level-deep reference file, and above 3 because structure and navigation are already good. | 4 / 5 |
Total | 15 / 20 Passed |