Conciseness

The body is command-dense and mostly lean — install, login, connection, and proxy sections are each a short command plus only the necessary state semantics. Minor trimmable spots: the content-free "Boost Overview" bullet tree ("Campaigns/Posts/Workspaces/Members" with no explanation), the duplicated action-discovery instructions in both "Searching for actions" and "Popular actions", and a stray docs link. This is anchor 4 (efficient, minor instances that could be trimmed), not anchor 5, which would require every section to earn its tokens.

Actionability

Nearly every section gives a copy-paste-ready command with flags: `membrane connection ensure "https://boostplatform.io/" --json`, `membrane action run <actionId> --connectionId=CONNECTION_ID --input '{"key": "value"}' --json`, and the headless-auth fallback `membrane login complete <code>`. It falls short of anchor 5 because commands use placeholders (CONNECTION_ID, QUERY, <actionId>) with no concrete real example of running an actual Boost action, and the "Popular actions" heading promises examples but delivers only a discovery command — plus the "Official docs" URL points to Microsoft Semantic Kernel, which is wrong.

Workflow Clarity

The connection flow has a genuine feedback loop: poll with `--wait`, branch on state (READY / CLIENT_ACTION_REQUIRED with clientAction sub-fields explained / CONFIGURATION_ERROR with "Check the `error` field"), and re-poll after user action — matching anchor 4's "clear sequence with most checkpoints present". Not anchor 5: the text says "skip to **Step 2**" twice and labels a section "1b" though no Step 1a/Step 2 headings exist (dangling navigation), CLI invocation is inconsistent (`membrane` vs `npx @membranehq/cli@latest`), and running actions has no error-handling guidance.

Progressive Disclosure

No bundle files exist, so scoring is on structure alone: the single ~140-line body has clean, well-ordered sections (Install → Authentication → Connecting → Searching/Running actions → Proxy → Best practices) that are easy to navigate, matching anchor 4's "good structure; most content is appropriately placed". It is not anchor 5 — reserved for skills under 50 lines with no need for external references — because the proxy flag table and clientAction state semantics are long enough that they could live in a one-level-deep reference file.

Specificity

"Boost integration. Manage data, records, and automate workflows" names the domain (Boost) but the actions are entirely generic — "manage data, records" and "automate workflows" could describe almost any SaaS skill, and no Boost-specific capabilities (scheduling posts, analytics, engagement, campaigns) are stated. It matches anchor 2 ("Names the domain but actions are minimal or generic") and falls short of anchor 3, which requires 1-2 concrete actions.

Completeness

Both elements are explicitly present: the "what" ("Boost integration. Manage data, records, and automate workflows") and an explicit trigger clause ("Use when the user wants to interact with Boost data"), so the missing-'Use when' cap of 3 does not apply and anchor 4 ("Has both 'what' and 'when'") is the best fit. It is not a 5 because the "what" is vague — no concrete capability phrases — and the "when" offers only a single trigger phrasing rather than concrete trigger variations.

Trigger Term Quality

The one genuinely relevant trigger term is the product name "Boost", repeated in "Use when the user wants to interact with Boost data" — but no natural variations or synonyms appear ("social media", "schedule posts", "analytics", "campaigns"). This matches anchor 3 (relevant keyword present, common variations missing); it is above anchor 2 because "Boost" is a term users of the platform would actually say, not pure generic filler.

Distinctiveness Conflict Risk

"Boost" is a distinctive product name, but the description never mentions social media, so "Boost integration" could plausibly trigger for unrelated Boost-named tools (e.g., the well-known C++ Boost libraries), and "Manage data, records, and automate workflows" overlaps with generic data-management skills. This is anchor 3 ("Somewhat specific but could still overlap with similar skills") — more overlap risk than anchor 4's "minor overlap risk with closely related skills".