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 practical, command-driven guide with a genuinely well-handled connection state machine (polling, client actions, error states) and copy-paste-ready CLI examples. Its weaknesses are padded framing content (the platform intro and hollow concept outline), a dangling "Step 2" reference, and no guidance on handling action-run errors.
Suggestions
Delete the "Boost.ai is a conversational AI platform..." intro paragraph and the concept-outline bullet list under "Boost.ai Overview", or replace them with one line mapping the domain concepts to the actions that manage them.
Define the missing "Step 2" (the actions workflow: search with "membrane action list --intent", then run) and renumber the connection steps consistently.
Add brief error-handling guidance for "membrane action run" (inspecting error output and retrying), mirroring the state-handling detail already given for connections.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Mostly efficient command-focused guidance, but it includes unnecessary padding Claude doesn't need: the opening paragraph explaining what Boost.ai is ("Boost.ai is a conversational AI platform that allows businesses to automate customer service..."), a hollow concept-outline bullet list ending in "Use action names and parameters as needed.", and prose like "This is the fastest way to get a connection. The URL is normalized to a domain...". It is not a 2 because the bulk is executable commands rather than library or concept explanation. | 3 / 5 |
Actionability | Concrete, executable commands throughout ("npm install -g @membranehq/cli@latest", "membrane connection ensure \"https://www.boost.ai/\" --json", "membrane action run <actionId> --connectionId=CONNECTION_ID --input '{\"key\": \"value\"}' --json") plus a flags table for proxy requests. It misses a 5 because "skip to **Step 2**" points to a step that is never defined, and invocation is inconsistent between "membrane" and "npx @membranehq/cli". | 4 / 5 |
Workflow Clarity | The connection sequence is clearly laid out with explicit state-based checkpoints and a polling feedback loop: poll while "BUILDING", branch on "CLIENT_ACTION_REQUIRED" (show uiUrl, then "poll again with membrane connection get <id> --json"), and stop on "CONFIGURATION_ERROR or SETUP_FAILED — check the error field". It misses a 5 due to the undefined "Step 2" reference and absent error-handling guidance for action runs. | 4 / 5 |
Progressive Disclosure | The skill is a single file with no bundle directories, and its content is organized into clear, well-labeled sections (Install the CLI, Authentication, Connecting to Boost.ai, Searching for actions, Best practices) with everything appropriately inline. It misses a 5 because at ~150 lines it exceeds the simple-skill threshold and contains noise sections — the concept-overview bullets and a "Popular actions" heading that lists no actions — instead of a lean overview. | 4 / 5 |
Total | 15 / 20 Passed |