Content
75%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 a strong, command-driven reference: every stage of the Membrane workflow has executable commands, explicit connection-state branching, and error-recovery guidance. Weaknesses are minor: a dangling 'Step 2'/'1b.' step numbering, an empty Description column in the actions table, and a little marketing filler at the top.
Suggestions
Fix the workflow numbering: the body references '**Step 2**' and uses a '#### 1b.' heading, but no Step 1 or Step 2 exists — either label the numbered steps or replace the references with section names like 'Searching for actions'.
Fill in or remove the empty Description column in the Popular actions table, and add one worked example with realistic inputs (e.g. membrane action run send-sms --input '{"to": "+15551234567", "message": "hi"}').
Trim the marketing intro ("It's used by organizations of all sizes...") and the filler line "Use action names and parameters as needed." to save context tokens.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dominated by lean, concrete commands, with only minor trimmable material: the three-sentence marketing intro ("It's used by organizations of all sizes to improve employee and customer communication"), the filler line "Use action names and parameters as needed.", and a Popular actions table with an entirely empty Description column. This matches 'efficient; minor instances of over-explanation that could be trimmed' — not 3 because the padding is a small fraction of the ~175 lines, and not 5 because those wasted tokens are real. | 4 / 5 |
Actionability | Commands are copy-paste ready (npm install, membrane login/complete, connection ensure with --wait polling, action list/run with --input, a complete proxy flag table) and error states are enumerated. It falls short of 'fully executable... specific examples cover the common cases' because there is no worked example with real inputs for a specific action (only the generic '{"key": "value"}') and the Popular actions table's empty Description column gives no guidance on what each listed action does. | 4 / 5 |
Workflow Clarity | The setup sequence is explicit with a genuine feedback loop: poll with --wait, branch on BUILDING/CLIENT_ACTION_REQUIRED (including clientAction.agentInstructions and uiUrl handling)/CONFIGURATION_ERROR, then re-poll after the user acts. One coherence gap prevents a 5: the body says "skip to **Step 2**" and has a "#### 1b." heading, but no Step 1 or Step 2 is labeled anywhere — those references dangle. | 4 / 5 |
Progressive Disclosure | No bundle files exist; this is a single-file skill of ~175 lines organized into clear sections (install → authenticate → connect → search → run → proxy → best practices), and the content is appropriately placed for that size. It matches 'good structure; most content appropriately placed; minor organization gaps' — the connection-state reference and proxy flag table are approaching a size where a separate reference file would help, keeping it below a 5. | 4 / 5 |
Total | 16 / 20 Passed |