Content
45%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 Membrane CLI sections (roughly the last 130 lines) are genuinely good — concrete commands, a well-handled connection state machine with polling, and sensible best practices. But they are buried under ~800 lines of machine-generated, massively repetitive entity and compliance-policy padding that wastes nearly the entire context budget and makes the skill hard to navigate. This is a strong 20% of a skill inside a broken 80% shell.
Suggestions
Delete the entire 'BlueCart API Overview' entity/compliance-policy bullet list (lines ~23-840) or collapse it to the handful of resources actually used (Product, Order, Category, Inventory); at minimum, deduplicate it — the same ~40 compliance entries repeat about ten times.
Move any needed resource/entity reference into a references/ file (e.g. references/resources.md) and keep SKILL.md as a concise overview pointing to it, per progressive disclosure.
Fix workflow numbering and consistency: label 'Step 2', reconcile the '1b' subsection, resolve where CONNECTION_ID comes from, and correct the 'Popular actions' table so descriptions match the BlueCart food/beverage framing (or update the framing to match the Walmart product-data API it actually wraps).
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The ~800-line 'Overview' bullet list is heavily padded filler, repeating the same ~40 compliance-policy entries (e.g. 'Modern Slavery Act Compliance Policy', 'Dodd-Frank... Compliance Policy') in near-identical loops roughly ten times over, and lists hundreds of entities with no instructional value. This squarely matches anchor 1 ('Severely verbose; heavily padded') — the padding dwarfs the genuinely useful material. | 1 / 5 |
Actionability | The working sections give concrete, executable commands: 'npm install -g @membranehq/cli@latest', 'membrane login --tenant --clientName=<agentType>', 'membrane connection ensure "https://www.bluecartapi.com/" --json', 'membrane action run <actionId> --input '{"key": "value"}' --json', plus a complete proxy flag table. Matches anchor 4 ('Mostly executable guidance with minor gaps'): placeholders like CONNECTION_ID are never resolved, the 'Agent Types ... etc.' line is vague, and the 'Popular actions' table is copy-pasted from a different skill's context. | 4 / 5 |
Workflow Clarity | There is a clear sequence (install CLI → login → connection ensure → poll state → search actions → run actions → proxy fallback) with real validation checkpoints: the connection state machine (READY / CLIENT_ACTION_REQUIRED / CONFIGURATION_ERROR) and re-polling after user action form a feedback loop. Not 5 because it references an unlabeled 'Step 2', has a '1b' subsection with no '1a', and never suggests verifying action output; matches anchor 4 ('Clear sequence with most checkpoints present; minor validation gaps'). | 4 / 5 |
Progressive Disclosure | No bundle files (references/, scripts/, assets/) exist, and the body inlines ~800 lines of entity/compliance-policy listing that clearly belongs in a separate reference file — worse than anchor 2's own 300-line example, though the working sections do have clear headers. Matches anchor 2 ('Minimal structure; content that clearly belongs in separate files is inlined'), not 3 because there are no references at all to signal. | 2 / 5 |
Total | 11 / 20 Passed |