octoquery: Skill for Claude Code

Blog database

The database is exposed as the MCP tool sql_blog_demo. Answer questions by sending a single SQL query string (MySQL dialect) to that tool. Results come back as JSON rows, truncated to the configured row limit (default 100) — use LIMIT, aggregation, or ORDER BY to keep result sets small and relevant.

Schema

• authors — id, email, full_name, joined_at
• posts — id, author_id → authors, title, category (DevOps, Databases, Programming, Web), status (draft, published, archived), views, published_at (NULL for drafts)
• comments — id, post_id → posts, commenter (free-text name, not a registered author), created_at, body

Relations

• One author has many posts (posts.author_id → authors.id).
• One post has many comments (comments.post_id → posts.id).
• Commenters are anonymous visitors — comments.commenter is just a name and does not reference the authors table.

Conventions

• Only published posts are publicly visible; exclude draft and usually archived posts from reader-facing metrics unless asked otherwise.
• views is a lifetime counter per post; published_at is NULL until a post is published.
• Popularity questions can mean views (posts.views) or engagement (comment counts) — pick from context or show both.

Guidelines

• Treat the database as read-only: stick to SELECT unless the user explicitly asks to modify data.
• Prefer one well-formed query over many small ones; use joins and aggregates server-side.