Conciseness

The body is mostly efficient (commands, flags table, actions table) but includes unnecessary explanation Claude does not need: "Capsule CRM is a customer relationship management (CRM) platform. It helps small to medium-sized businesses manage contacts, sales pipelines..." and a content-free entity list ("**Opportunity**, **Track**, **Case**...") with no descriptions, plus prose like "This is the fastest way to get a connection. The URL is normalized to a domain...". This matches anchor 3 (mostly efficient, some unnecessary explanation) rather than anchor 4, where only minor trimming would be needed — the intro paragraph and entity list add no actionable value.

Actionability

Every step ships copy-paste-ready commands: "npm install -g @membranehq/cli@latest", "membrane login --tenant --clientName=<agentType>", "membrane connection ensure \"https://capsulecrm.com/\" --json", "membrane action run <actionId> --connectionId=CONNECTION_ID --input '{\"key\": \"value\"}' --json", and a complete flag table for proxy requests. The 20-row popular-actions table covers the common cases with concrete keys, matching anchor 5; it is not lower because no step relies on pseudocode or vague direction.

Workflow Clarity

A sequence exists (install → login → connection ensure → wait for READY → search actions → run → proxy fallback) with state-based branching for CLIENT_ACTION_REQUIRED, but it hits the rubric's destructive-operation cap: the actions catalog includes "Delete Project" and "Delete Task" yet no validation or confirmation guidance exists for destructive operations (no verify-the-ID / confirm-with-user step). It also matches anchor 3 rather than 4 due to an incoherent reference — "skip to **Step 2**" points at sections that are never numbered, leaving checkpoints implicit.

Progressive Disclosure

The single SKILL.md (no bundle files present) is well-organized into clear sections (Install, Authentication, Connecting, Searching, Running, Proxy, Best practices) with the actions table and flag table appropriately inline as the core discovery surface, and only an external docs URL as a reference. It is not 5 because the >50-line body keeps material like the full 20-row actions table and proxy flag table inline where a reference file split could apply, and stray structure issues ("1b" heading with no "1", the "Step 2" reference) slightly hurt navigation — but these are minor gaps consistent with anchor 4.

Specificity

"Manage crm and sales data, records, and workflows" names the domain (Capsule CRM) but every action is a generic verb on generic objects — no concrete operations like creating contacts, tracking opportunities, or managing tasks. This matches anchor 2 ("Names the domain but actions are minimal or generic") rather than anchor 3, which requires 1-2 concrete actions akin to "extracts content".

Completeness

Both parts are present: the "what" ("Manage crm and sales data, records, and workflows") and an explicit "Use when the user wants to interact with Capsule CRM data" trigger. It is not a 5 because the "when" clause is narrow (only "interact with... data") and the "what" leans on generic verbs, so neither is fully concrete.

Trigger Term Quality

Terms "Capsule CRM", "crm and sales data", "records", and "workflows" give good natural-keyword coverage of the domain. It falls short of anchor 5 because common CRM phrasings users actually say — contacts, leads, opportunities, deals, sales pipeline — are absent.

Distinctiveness Conflict Risk

"Capsule CRM" is a specific product name appearing in both the what and when clauses, giving a clear niche with distinct triggers and minimal conflict risk against generic CRM/sales skills. The rubric's third-person voice requirement is met ("Manage..."), so no penalty applies.