Content
61%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, largely copy-paste-ready CLI guide with a well-sequenced connection workflow and explicit state-handling checkpoints. Its main flaws are padding (the empty Overview section and intro paragraph), a dangling 'Step 2' reference that breaks the numbered flow, and no worked example instantiating the placeholder IDs.
Suggestions
Delete the content-free 'Carbon Ads Overview' bullet list and trim the intro paragraph about what Carbon Ads is — Claude already knows this, and the bullets carry no information.
Fix the dangling 'skip to Step 2' references by actually labeling the subsequent sections as Steps 1 and 2 (or removing the numbering), since no Step 2 is ever defined.
Add one fully instantiated worked example (e.g. listing and running a campaign action with a real connection id and --input payload) and a brief error-check step after 'membrane action run' to close the actionability and workflow-validation gaps.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The bulk is an efficient command reference, but there is unnecessary padding: an opening paragraph explaining what Carbon Ads is, a content-free 'Carbon Ads Overview' bullet list (Campaign/Creatives/Account with no descriptions), and loose prose like 'This will burn less tokens'. Fits anchor 3 ('mostly efficient but includes some unnecessary explanation or could be tightened'); not 4 because the empty Overview section and intro explanation are clearly trimmable. | 3 / 5 |
Actionability | Commands are concrete and executable throughout: 'npm install -g @membranehq/cli@latest', 'membrane connection ensure "https://carbonads.net" --json', 'membrane action run <actionId> --connectionId=CONNECTION_ID --input ...', plus a full proxy options table. Not 5 because placeholders like CONNECTION_ID/actionId are never instantiated in a worked example and some guidance stays abstract. | 4 / 5 |
Workflow Clarity | The sequence (install → login → connection ensure → poll for READY → discover actions → run → proxy fallback) is clear, with state-based checkpoints (READY/BUILDING/CLIENT_ACTION_REQUIRED) and a re-poll loop after user action. Not 5 because 'skip to Step 2' references a Step 2 that is never defined in the document, and there are no validation/error-handling steps for action runs themselves. | 4 / 5 |
Progressive Disclosure | The skill is a single self-contained file (no references/ or bundle files exist to verify), organized into clear headed sections (Install, Authentication, Connecting, Searching, Running, Proxy, Best practices) with no nested references. Not 5 because at ~130 lines, material like the proxy flag table and action discovery detail sits inline where a small reference file could keep SKILL.md leaner. | 4 / 5 |
Total | 15 / 20 Passed |