Content
82%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 strongly actionable integration guide: copy-paste-ready Membrane CLI commands for auth, connection setup with state-machine handling, action discovery/execution, and proxy fallback. Main weaknesses are the broken step references ('Step 2', '1b' with no matching labels), the low-value bare entity list, and no validation guidance for action results or destructive operations.
Suggestions
Fix the incoherent step numbering: either label the sections 'Step 1'/'Step 2' consistently or replace 'skip to **Step 2**' and '#### 1b.' with descriptive headings.
Cut or enrich the 27-item bare entity list ('**Matter**', '**Note**', ...) — either move it to a reference file or annotate entries with what each object is for; also delete the filler line 'Use action names and parameters as needed.'
Add a validation step after running actions — e.g. check the output/error fields of 'membrane action run' before proceeding, especially before destructive actions like delete-lead.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is mostly lean, executable command blocks and tables with no over-explanation of concepts Claude already knows. The 27-item entity list under 'CINC Overview' (e.g. '**Matter**', '**Note**') is bare names with no accompanying guidance and reads as padding, plus the filler line 'Use action names and parameters as needed.' — minor trimmable instances matching score 4 rather than the fully lean score 5. | 4 / 5 |
Actionability | Every section gives copy-paste-ready commands with exact flags: 'npm install -g @membranehq/cli@latest', 'membrane connection ensure "https://www.cincpro.com/" --json', 'membrane action run <actionId> --connectionId=CONNECTION_ID --input ... --json', plus a flags table for proxy requests and a 14-row popular-actions table covering common cases. It is fully executable and covers the common cases, matching the score 5 anchor. | 5 / 5 |
Workflow Clarity | The sequence (install → login incl. headless flow → connect → poll state → search → run) is clear, with a genuine feedback loop for connection states (READY / CLIENT_ACTION_REQUIRED / CONFIGURATION_ERROR with clientAction fields to inspect). It falls short of 5 due to incoherent step numbering — 'skip to **Step 2**' and '#### 1b.' reference steps that are never labeled elsewhere — and no validation guidance for action-run results or destructive actions like delete-lead, which are minor gaps fitting score 4's 'most checkpoints present'. | 4 / 5 |
Progressive Disclosure | No bundle files exist, so all content is necessarily inline in SKILL.md; sections are clearly headed and easy to navigate (CLI setup, connection, actions, proxy, best practices). The bare entity list and full actions table are content that could live in a separate reference file if a bundle were used, which keeps it at 'Good structure; most content appropriately placed; minor organization gaps' rather than 5. | 4 / 5 |
Total | 17 / 20 Passed |