/integrations — Third-party integration page generator
Integration pages capture a narrow but high-converting search behavior: prospects who already want the site's product but need to verify it plays nicely with their existing stack. These pages also stack topical authority — a SaaS with 50 integration pages outranks one with 0 even if the underlying product is identical.
This skill is most useful for SaaS, API products, and developer tools. Service businesses (consulting, catering, medical) rarely need integration pages — the skill flags this at start.
Inputs
Required (asked at start if not given):
- Site — must match a folder under
sites/[name]/. Must haveproduct_type=saasorproduct_type=apiinsite-info.md. - Integration target — the named tool (Salesforce, HubSpot, Zapier, Slack, Notion, etc.).
Optional (per-run overrides):
- Integration type —
native(built-in feature) |via-zapier(no-code middleware) |via-api(custom API connection) |via-webhook(event-driven). Defaults to reading from site's integrations registry insite-info.md. - Verified by —
engineering(defaults — implies actual integration code exists) |roadmap(planned but not shipped — flagged on page). Skill refuses if undeclared.
Workflow
This skill follows /service's workflow (Steps 0–11) with the page-type-specific differences listed below. Read .claude/skills/service/SKILL.md end-to-end before running.
What's unique to /integrations
Step 0 — Product-type guard: if site-info.md does not declare product_type=saas or product_type=api, the skill warns and asks for confirmation. Service-business sites rarely benefit from integration pages.
Step 2 — Keyword selection:
- Common forms:
[Site] [Tool] integration,[Site] + [Tool],does [Site] work with [Tool]?,connect [Site] to [Tool]. - Pick from
service-keywords.csvrows with intent=integration. If none exist, add one before proceeding.
Step 4 — SERP analysis (integration-specific):
- Top 3 results — these are often the target tool's own integration directory (Zapier's Salesforce page, HubSpot's marketplace). Hard to beat for the exact query. Aim to rank for the long-tail variant ("does [Site] sync [field] to [Tool]?") rather than the head term.
- Extract: do the top results show real screenshots? Real config steps? Real field-mapping tables? Most do — match or exceed.
Step 5 — Integration data gathering:
- Required from engineering (or
site-info.mdintegrations registry):- What data flows (fields, frequency, direction — one-way vs two-way sync)
- Configuration steps (numbered, from zero to working)
- Authentication method (OAuth, API key, webhook URL)
- Known limitations (rate limits, unsupported fields, edge cases)
- Pricing implications (does this integration cost extra? require a higher plan?)
If ANY of the 5 is missing from documented sources, refuse — integration pages that fudge details rank temporarily but get rebuked in reviews and damage trust.
Step 6 — Wireframe (integration-specific):
- Hero: "Connect [Site] to [Tool] in [time-estimate] minutes" or "[Site] + [Tool]: the [primary-use-case] integration"
- Quick answer (above fold): 2-sentence summary — "What syncs: [primary-fields]. How: [native | Zapier | API]. Time to set up: [minutes]."
- Use cases section (3–5 named scenarios): each 80–150 words. "If you're a [role] in [team-type], here's what this unlocks for [workflow]."
- Data flow diagram or table (required): which fields move in which direction, at what frequency, under what trigger. Visual or table — pick based on complexity.
- Configuration steps (numbered, screenshots): the actual zero-to-working setup. This is the section that wins the SERP — most integration pages skip the real steps.
- Limitations section (required, not optional): rate limits, unsupported fields, gotchas. Trust signal.
- Pricing note (if applicable): "This integration is included in [plan]" or "Requires [plan] starting at $X."
- FAQ: 5–8 Qs — "Is this real-time or batch?", "Does it support [edge-case-field]?", "Can I customize the mapping?", "What if my [Tool] account uses [variant]?"
- CTA: "Start a free trial" (for SaaS) or "Try the integration" (API products).
Step 7 — Schema (integration-specific):
ServicewithserviceType: "Software integration"andprovider: [our org]mentionsfor the integrated tool's organization (e.g.,{@type: SoftwareApplication, name: "Salesforce"})HowToschema for the configuration steps (each step as aHowToStep)FAQPageBreadcrumbList
Step 8 — Voice (integration-specific extras):
- Every claim about the integration's capabilities MUST be in
site-info.mdintegrations registry or in supplied dossier. No "we plan to" implied as "we have." - If integration is
verified-by=roadmap, the page MUST show a banner: "This integration is on our roadmap. Estimated availability: [date]." Page can still exist for SEO capture, but the banner is non-removable. - One opinion max — usually "[Use case] is where this integration matters most" backed by usage data from stats.md.
- AI-search rules: Q+A density critical — integration questions are heavily LLM-cited ("does Acme work with Salesforce?" is a common LLM-mediated query).
Refusal conditions
This skill refuses to ship if:
- Site is not
product_type=saasorproduct_type=api. - Any of the 5 required integration data points is undocumented.
- Integration is
verified-by=roadmapbut the date estimate is missing. - An integration page for this tool already exists in
used-keywords.md(refresh via/refresh).
Output
Standard _drafts/[slug]/ layout. Slug pattern: integrations/{tool-slug} (path-based to support the multi-page integration directory pattern).
Length target: 800–1,800 words. Integration pages skew shorter than blog posts — readers want config and limitations, not narrative.
Routine versioning + idempotency contract
Same as /service. Idempotency key: (site, tool-slug, language). Integration pages auto-decay — refresh quarterly when the integration registry version-bumps.