Knowledge Base

Plan a knowledge base index with named owners, tags and review dates. The skill starts with existing content and improves findability before recommending new writing. It produces empty structures for documentation records on request.

The skill separates advice, review, conversion and build requests before producing artifacts. Advice receives a short recommendation and an offer of a relevant build. An explicit build request proceeds once essential missing facts are resolved. It asks one question at a time, reuses answers already supplied and stops when further answers would not change the deliverable. Unknown information remains unknown rather than receiving invented values.

Use it for company wiki articles, a help center database or an internal documentation index. Clarify current storage, search habits, audience, content owners and review cycles.

It does not write the content or replace a full wiki platform. A named owner and a review date per page are the minimum; creating missing articles follows the index and findability work. The skill does not calculate payroll, tax, leave balances or KPIs. Automation, reminders and synchronization require an integration layer. Legal, tax and HR review is required before templates drive real decisions.

Inputs cover document types and volume, authors, tags or folders, internal visibility and how outdated content is removed. The workflow captures content, assigns an owner, adds tags and structure, supports search and establishes a review cycle. Its 12 fields are Article Title, Category, Department, Owner, Tags, Summary, Linked SOP, Audience, Last Reviewed, Next Review, Status and Article ID. Linked SOP is the relation field. Categories include Process, How To, Policy, Template, Reference and FAQ; statuses are Draft, In Review, Published and Archived.

Requested artifacts can include CSV, SQL DDL, JSON Schema, Notion property mappings or an Excel workbook. The formats derive from one field list: CSV and JSON retain matching display names, while SQL uses normalized identifiers. Dates retain date types, money retains currency types and cross-table references retain relation types. Select options are a starting set that can be renamed to match the business.

One SQL example is:

CREATE INDEX idx_knowledge_base_status ON knowledge_base (status);

Templates are empty by default. Example rows are produced only on request and remain obviously synthetic. Sensitive identifiers should not be repeated unnecessarily during a requested review. Machine-readable data stays separate from explanatory prose and any material validation failures. Excel-compatible CSV uses UTF-8 with a byte order mark and accompanying guidance about number, date and currency columns. An XLSX workbook is a separate requested output rather than a renamed CSV.

The Notion output routes to notion-manual-import for CSV, property mappings, import steps and verification. Manual mappings require no connection. Requested live workspace changes require actual tool access, a verified target and authorization. Imported columns may initially be Text and need their mapped property types applied. Relation links resolve after both relevant databases are imported. A static mapping alone does not establish a completed workspace change.