Content
63%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 solid, executable Membrane operational guide with a well-handled connection state machine and a sensible discover-before-build philosophy. Its weaknesses are an unnecessary vendor background intro, dangling 'Step 2' cross-references, an empty Description column in the actions table, and no offloading of reference material to bundle files.
Suggestions
Delete the "Anthropic is an AI safety and research company..." intro paragraph and the vague "Anthropic Overview" bullet list — Claude already knows this, and the space is better spent on a worked `membrane action run` example with real parameters.
Fix the step cross-references: either label the sections "Step 1 / Step 2" so "skip to **Step 2**" and the "1b" heading resolve, or reword to plain section names.
Fill in the empty Description column of the Popular actions table (or link to a reference file listing all actions with their input schemas), and consider moving the proxy flags table to a `references/` file to keep SKILL.md a lean overview.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The bulk is lean, copy-pasteable CLI instruction, but there is padding Claude does not need: the opening paragraph explaining what Anthropic and Claude are ("Anthropic is an AI safety and research company... competitor to OpenAI's GPT models") and the vague "Anthropic Overview" bullets ending in "Use action names and parameters as needed." This fits 'Mostly efficient but includes some unnecessary explanation or could be tightened' rather than 2, which requires several padded sections throughout. | 3 / 5 |
Actionability | Commands are concrete and executable throughout — "npm install -g @membranehq/cli@latest", "membrane connection ensure \"https://console.anthropic.com\" --json", "membrane action list --connectionId=... --intent ... --json", plus a complete proxy flag table. Minor gaps keep it at 4 rather than 5: the Popular actions table has an entirely empty Description column, and running actions is shown only with placeholder inputs ('{"key": "value"}') rather than a worked example with real parameter values. | 4 / 5 |
Workflow Clarity | The install → authenticate → connect → poll → discover → run sequence is clear, with explicit state-machine checkpoints (READY / CLIENT_ACTION_REQUIRED / CONFIGURATION_ERROR) and a feedback loop of re-polling after the user completes an action. Minor gap: the text says "skip to **Step 2**" and defines a "1b" step, but no Step 1 or Step 2 headings exist, leaving the numbered sequence dangling. | 4 / 5 |
Progressive Disclosure | A single ~150-line file with sensible section headers and no external references; at this size it exceeds the under-50-line simple-skill exception, and content that could live in separate files (the proxy options table, the full actions listing) is inlined. This matches 'Some structure but could be better organized' — above 2 because the structure is real and navigable, below 4 because nothing is split out or referenced. | 3 / 5 |
Total | 14 / 20 Passed |