Content
67%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, executable runbook for the Membrane-mediated Billy integration, with real command examples and state-based polling checkpoints. Its main flaws are leftover placeholder text, a duplicated/empty "Popular actions" section that never names a single action, and broken internal step references.
Suggestions
Delete the placeholder line "I don't have enough information about that app to provide a description." and replace the empty Billy Overview bullet list with the actual action IDs or intents for those entities (contacts, companies, deals, activities, tasks, emails).
Make "Popular actions" actually list 5-10 concrete example actions with their IDs or intents, or merge it into "Searching for actions" to remove the duplicated `action list` command.
Fix the dangling workflow references: either add the numbered "Step 2" (e.g. searching for actions) that "skip to Step 2" points to, or relabel the sections as a coherent numbered sequence instead of the orphaned "1b".
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Mostly efficient — the bulk is lean CLI examples with brief, useful explanations — but several things could be trimmed: the placeholder line "I don't have enough information about that app to provide a description.", the empty nested "Billy Overview" bullet list, and the action-discovery command duplicated across "Searching for actions" and "Popular actions". Fits anchor 3 ("mostly efficient but includes some unnecessary explanation or could be tightened"); not 4 because these are more than minor instances. | 3 / 5 |
Actionability | Concrete, copy-paste-ready commands throughout: `npm install -g @membranehq/cli@latest`, `membrane login`, `membrane connection ensure`, `membrane action run <actionId> --input '{...}'`, plus a proxy flag table. Falls short of 5 because the "Popular actions" section never names an actual action and the overview's entity categories are never connected to specific action IDs. | 4 / 5 |
Workflow Clarity | Clear install → login → connect → poll → discover → run sequence with explicit state-based checkpoints (READY, BUILDING, CLIENT_ACTION_REQUIRED, CONFIGURATION_ERROR) and a re-poll loop after the user completes an action. Held below 5 by a broken "skip to **Step 2**" reference (no Step 2 exists) and inconsistent step numbering ("1b" with no labeled 1a/2). | 4 / 5 |
Progressive Disclosure | A single-file skill with no bundle files, well-sectioned headers, and dynamic content (actions discovered at runtime via `action list`) so there is little that belongs in separate files. Good structure with minor gaps: odd step numbering, the redundant "Popular actions" section, and a dead overview list. | 4 / 5 |
Total | 15 / 20 Passed |