Content
57%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 genuinely executable Membrane CLI instructions and a decent connection-readiness feedback loop. Weaknesses are a dangling "Step 2" reference and unnumbered step labels, an empty "Popular actions" section and content-free overview, and inlined generic Membrane material that inflates the single-file body.
Suggestions
Fix the broken cross-references: replace "skip to **Step 2**" with the actual section name (e.g., "### Searching for actions") and either number the steps consistently (1, 1b, 2) or drop the numbering entirely.
Either populate "Popular actions" with real Blend action examples (actionId + sample --input) or remove the empty section and the filler line "Use action names and parameters as needed."
Move the generic Membrane CLI setup (login, connection polling, proxy flags) into a references/ file (e.g., membrane-cli.md) and keep SKILL.md focused on Blend-specific usage, which would also improve token efficiency.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Mostly efficient — the body is dominated by concrete commands — but several instances waste tokens: the "Blend Overview" hierarchy (Document/Page, Workspace/User) carries no information, "Use action names and parameters as needed." is pure filler, and "Those will be used to adjust tooling to be used best with your harness" is garbled padding. Not 4 because these are more than minor trims. | 3 / 5 |
Actionability | Commands are concrete and executable throughout: `npm install -g @membranehq/cli@latest`, `membrane connection ensure "https://blend.com/" --json`, action run with `--input` JSON payloads, and a full proxy flag table. Minor gaps: no example uses a real Blend actionId, and the "Popular actions" section lists no actual actions. | 4 / 5 |
Workflow Clarity | A real sequence exists (install → login → connection ensure → poll for READY → search actions → run) with a connection-state feedback loop (READY / CLIENT_ACTION_REQUIRED / CONFIGURATION_ERROR), but "skip to **Step 2**" references a step that does not exist and "1b" appears with no "1a". No error-handling guidance exists for running actions. Not 4 because the dangling step reference breaks sequence coherence. | 3 / 5 |
Progressive Disclosure | Well-sectioned single file, but at ~140 lines with generic Membrane CLI setup/auth/proxy material inlined that is common to any Membrane-based skill and would belong in a shared reference. No bundle files exist, so all content sits in SKILL.md with only an external docs URL. Some structure, but content that should be separate is inline. | 3 / 5 |
Total | 13 / 20 Passed |