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's operational core is genuinely strong — every Membrane workflow step has a copy-pasteable command and connection-state edge cases are handled thoroughly. Its weaknesses are dead weight and dead references: a contentless overview outline, a 20-row table of "No description" entries, a dangling "Step 2" pointer, and no validation guidance for write operations. Trimming the filler into a reference file and fixing the step numbering would move this from adequate to strong.
Suggestions
Fix the dangling reference "skip to **Step 2**" — either number the sections (Step 1: Connect, Step 2: Find and run actions) or reword to "skip to \u2018Searching for actions\u2019", the lowest-hanging workflow-clarity fix.
Add a validation step for write operations (e.g. after update-deal/update-person, re-run the corresponding get-* action to confirm the change) — this would also lift the workflow-clarity cap on destructive/batch operations.
Replace the "No description" cells in the Popular actions table with one-line descriptions and input hints, or move the full table to a references/actions.md file and keep only the top few inline — this addresses both conciseness and progressive disclosure.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The bulk of the body is lean, executable CLI material (install, login, connection, action, proxy commands and flag tables), but it carries clear padding: an empty "Agendor Overview" outline ("Contact / Task / Company / Task" plus "Use action names and parameters as needed"), a 20-row Popular actions table where every Description cell reads "No description", and an introductory paragraph explaining what a CRM is. It sits between anchors 2 and 3 — several padded sections exist, but they are a minority of the document. | 3 / 5 |
Actionability | Concrete, executable commands appear for every phase: "npm install -g @membranehq/cli@latest", "membrane connection ensure \"https://www.agendor.com.br/\" --json", "membrane action list --connectionId=CONNECTION_ID --intent \"QUERY\"", and a parameterized run example with --input. It misses anchor 5 because there is no end-to-end worked example using a real action key (e.g. list-deals) with actual parameters, and the actions table gives no input-schema guidance. | 4 / 5 |
Workflow Clarity | A real sequence exists (install → login → connection ensure → poll until READY → search actions → run), with a good feedback loop for connection states (BUILDING polling, CLIENT_ACTION_REQUIRED handling, CONFIGURATION_ERROR/SETUP_FAILED). However, the text says "skip to **Step 2**" while no section is labeled Step 2 — a broken cross-reference — and there is no validation/verification guidance for write operations (update-deal, update-person, arbitrary proxy PUT/POST/DELETE), which caps workflow clarity at 3 per the destructive/batch validation rule. | 3 / 5 |
Progressive Disclosure | The skill ships no bundle files (no references/, scripts/, or assets/ exist) and the body is a single ~160-line file. Section headers are clear, but bulk reference material — the 20-row actions table and the proxy flag table — is inlined in SKILL.md where it belongs in a separate reference file, matching anchor 3 ("content that should be separate is inline"). It is not anchor 2 because headers and navigation are present and coherent. | 3 / 5 |
Total | 13 / 20 Passed |