Content
63%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 genuinely actionable — executable CLI commands with flags, a state-machine walkthrough for connection setup, and concrete error-handling guidance — held back by dead weight and loose structure. The placeholder 'Clari Overview', the duplicated action-discovery command, the dangling 'Step 2' reference, and the absence of any reference files for the bulkier detail (proxy flags, clientAction schema) are the main fixable gaps.
Suggestions
Delete the placeholder 'Clari Overview' bullet list and the duplicated 'Popular actions' section (it repeats the 'Searching for actions' command); replace the former with a short mapping of Clari entities (Opportunity, Account, User, Activity) to example action names, and the latter with 5-10 concrete popular actions with their IDs.
Fix the broken step numbering: either label the sections 'Step 1: Connect', 'Step 2: Find actions', 'Step 3: Run actions' so 'skip to Step 2' resolves, or drop the numbered references entirely.
Move the proxy flag table and the clientAction field schema into a reference file (e.g. references/proxy.md) with a clearly signaled link, keeping SKILL.md as a lean overview — this also addresses the duplicated install of detail that inflates the body.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is mostly command-plus-one-line-explanation, but it carries clear waste: the marketing-style opener ('helps sales, marketing, and customer success teams improve their forecasting...'), a 'Clari Overview' section that is an empty placeholder bullet list ('Opportunity / User / Activity / Account / User / User' plus 'Use action names and parameters as needed.'), and a 'Popular actions' section that duplicates the 'Searching for actions' command verbatim. Anchor 3 ('mostly efficient but includes some unnecessary explanation or could be tightened') fits; the duplicated and dead sections rule out 4. | 3 / 5 |
Actionability | Nearly every section gives copy-paste-ready commands: 'npm install -g @membranehq/cli@latest', 'membrane connection ensure "https://clari.com" --json', 'membrane action run <actionId> --connectionId=CONNECTION_ID --input \'{"key": "value"}\' --json', plus a concrete proxy flag table. Minor gaps keep it at anchor 4 rather than 5: 'membrane login --tenant --clientName=<agentType>' shows --tenant with no value, and the 'Popular actions' heading promises actions but only repeats the discovery command instead of naming any. | 4 / 5 |
Workflow Clarity | The connection flow has real checkpoints and a feedback loop — poll with '--wait', branch on 'READY' / 'CLIENT_ACTION_REQUIRED' (with the clientAction type, uiUrl, and agentInstructions fields explained) / 'CONFIGURATION_ERROR' or 'SETUP_FAILED', then re-poll after the user acts — matching anchor 4's 'clear sequence with most checkpoints present'. Not 5 (and not the simple-skill exception) because the sequence is partially incoherent: 'skip to Step 2' points at a step that is never labeled, and the lone '#### 1b' heading references a numbering scheme (1a/2) that doesn't exist. | 4 / 5 |
Progressive Disclosure | There are no bundle files at all — references/, scripts/, and assets/ do not exist — and all content (install, auth, connection lifecycle, clientAction schema walkthrough, proxy flag table) is inline in a single ~140-line file with usable headers. Anchor 3 ('some structure but could be better organized; content that should be separate is inline') fits: the proxy flag table and clientAction field details are natural candidates for one-level-deep reference files. The under-50-line exception does not apply, ruling out 5; decent sectioning rules out 2. | 3 / 5 |
Total | 14 / 20 Passed |