Content
76%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.
This is a highly actionable, command-driven integration skill: every workflow step ships a copy-paste membrane command, and the connection lifecycle (including BUILDING/CLIENT_ACTION_REQUIRED states) is handled thoroughly. Its main weaknesses are a broken 'Step 2' cross-reference with unnumbered steps, no validation guidance for destructive actions like deletes, and a few content-free overview sections that pad the token budget.
Suggestions
Fix the dangling 'skip to **Step 2**' reference by numbering the workflow steps explicitly (1. Connect, 2. Search actions, 3. Run action) so the '1b' subsection and forward references resolve.
Add a validation checkpoint for destructive actions: confirm the record ID with the user (or run get-deal/get-contact first) before running delete-deal/delete-company/delete-contact.
Delete the contentless 'Clientify Overview' bullet list ('Company > Contact / Deal / Use action names and parameters as needed') or replace it with what the entities mean for action selection.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dominated by tight, copy-paste command blocks and a compact actions table, with only minor trims needed: the marketing-style intro ('Clientify is an all-in-one business management platform...'), the contentless 'Clientify Overview' section ending in 'Use action names and parameters as needed', and the parenthetical Agent Types aside. This fits 'Efficient; minor instances of over-explanation that could be trimmed' rather than 5, where every token would earn its place. | 4 / 5 |
Actionability | Every stage has fully executable commands: 'npm install -g @membranehq/cli@latest', 'membrane connection ensure "https://clientify.com/" --json', 'membrane action run <actionId> --connectionId=CONNECTION_ID --input ...', plus a flags table for proxy requests and a 20-row actions table with keys. Specific examples cover the common cases (headless login, waiting on BUILDING state, searching by intent), matching the top anchor; the only placeholders are legitimately variable values. | 5 / 5 |
Workflow Clarity | The sequence (install → login → connection ensure → handle state → search actions → run → proxy fallback) is present, and connection state handling forms a genuine feedback loop, but the text says 'skip to **Step 2**' when no numbered Step 2 section exists (steps are never numbered; there is only an unanchored '1b'), and destructive operations (delete-deal, delete-company, delete-contact) have no verification or confirmation guidance, capping this dimension at 3 per the destructive/batch rule. | 3 / 5 |
Progressive Disclosure | As a single-file skill with no bundle files (references/, scripts/, assets/ are absent), it is well organized into clear sections (Install, Authentication, Connecting, Searching, Popular actions, Running, Proxy, Best practices). Minor gaps keep it below 5: the 20-row actions table and the detailed connection-state reference are inline content that could live in a references file if the skill grows, and it is above the under-50-line threshold for the simple-skill exception. | 4 / 5 |
Total | 16 / 20 Passed |