Content
71%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 skill is highly actionable with a well-instrumented connection workflow and copy-paste commands. Its main weakness is token efficiency: a large redundant endpoint catalog is inlined in SKILL.md rather than externalized into a reference file.
Suggestions
Move the endpoint catalog into a separate reference file (e.g. references/endpoints.md) and keep SKILL.md as an overview, since the same data is discoverable via `membrane action list --intent`.
Trim the redundant 'Get X — Retrieves X' endpoint list, which restates each name as its description and overlaps the Popular actions table.
Add brief guidance on interpreting and validating `membrane action run` output (e.g. checking the `output` field and handling error states) to complete the workflow's feedback loop.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The Membrane CLI instructions are efficient, but the ~85-line endpoint catalog restates each endpoint's name as its description ('Get Featured Properties — Retrieves a list of featured properties') and overlaps the Popular actions table, which is padded content Claude could instead discover via `membrane action list`. | 3 / 5 |
Actionability | Fully executable, copy-paste-ready commands cover the common cases end-to-end — install, login (including headless completion), connection ensure/wait, action list, action run with JSON input, and proxy requests — each with the relevant flags. | 5 / 5 |
Workflow Clarity | The connection flow is a clear sequence with explicit state checkpoints (READY/BUILDING/CLIENT_ACTION_REQUIRED/CONFIGURATION_ERROR) and a poll-and-retry feedback loop; the minor gap is the absence of guidance on validating or handling `action run` results and errors. | 4 / 5 |
Progressive Disclosure | Section headers and a Popular actions table provide some structure, but the full endpoint catalog is inlined in SKILL.md with no bundle/reference files present, so content that belongs in a separate one-level-deep reference is not split out. | 3 / 5 |
Total | 15 / 20 Passed |