Starter Writing skill
Build a customer FAQ
Create a customer FAQ from recurring evidence with concise answers, exceptions, actions, ownership, and review dates.
Build around questions customers actually ask, not questions the organization wishes they asked. Each answer should resolve one intent and point to a safe action or responsible contact.
Inputs
- Gather anonymized support themes, site-search terms, onboarding questions, policy pages, and product documentation.
- Confirm the audience, publication surface, current owners, and source of truth for each answer.
- Rank candidate questions by frequency, impact, confusion, and self-service suitability.
Procedure
- Cluster equivalent questions while preserving important differences in user intent.
- Select a bounded first set and write each heading in the customer's language.
- Answer directly in the first sentence, then add necessary steps, limits, or examples.
- Link to the canonical action, policy, or detailed guide and verify the destination.
- State exceptions and escalation paths where a general answer is not sufficient.
- Remove internal terminology, promotional filler, unsupported claims, and duplicate answers.
- Assign an owner, source, and review trigger for every answer.
- Test the page with real questions, keyboard navigation, mobile layout, and site search.
- Measure searches, exits, unresolved contacts, and feedback without treating fewer tickets as the only success signal.
Boundaries
Never expose customer examples, account data, internal security procedures, or private policy exceptions. Do not present legal, medical, tax, or financial guidance beyond approved language. Keep a human support path for cases that require context or consent.
--- name: build-a-customer-faq category: write description: Create a customer FAQ from recurring evidence with concise answers, exceptions, actions, ownership, and review dates. Use when support questions repeat and customers need a trustworthy self-service page. --- # build-a-customer-faq Build around questions customers actually ask, not questions the organization wishes they asked. Each answer should resolve one intent and point to a safe action or responsible contact. ## Inputs - Gather anonymized support themes, site-search terms, onboarding questions, policy pages, and product documentation. - Confirm the audience, publication surface, current owners, and source of truth for each answer. - Rank candidate questions by frequency, impact, confusion, and self-service suitability. ## Procedure 1. Cluster equivalent questions while preserving important differences in user intent. 2. Select a bounded first set and write each heading in the customer's language. 3. Answer directly in the first sentence, then add necessary steps, limits, or examples. 4. Link to the canonical action, policy, or detailed guide and verify the destination. 5. State exceptions and escalation paths where a general answer is not sufficient. 6. Remove internal terminology, promotional filler, unsupported claims, and duplicate answers. 7. Assign an owner, source, and review trigger for every answer. 8. Test the page with real questions, keyboard navigation, mobile layout, and site search. 9. Measure searches, exits, unresolved contacts, and feedback without treating fewer tickets as the only success signal. ## Boundaries Never expose customer examples, account data, internal security procedures, or private policy exceptions. Do not present legal, medical, tax, or financial guidance beyond approved language. Keep a human support path for cases that require context or consent. ## Done - Every published question traces to observed customer evidence - Answers are accurate, concise, actionable, and checked against current sources - Exceptions, escalation routes, owners, and review triggers are written - Representative customer questions find the correct answer in search and usability checks