Agent Skills: Scott Blog Post

Scott Blog Post

Use this skill for scottspence.com posts.

Quick Start

Build evidence first, choose a high-intent slug/title, draft into posts/<slug>.md, then validate with Prettier, cspell, and preferably pnpm build.

Workflow

1. Read nearby/relevant posts from posts/.
2. Query data/site-data.db with mcp-sqlite-tools for recent slugs, titles, tags, and positioning.
3. Inspect referenced repos/files and recent git churn before making claims.
4. Use pirecall/ccrecall or mcp-sqlite-tools when the story comes from session history.
5. Keep the prose evidence-led, personal, useful, and free of generic AI-blog framing.

Voice rules

• Conversational, blunt, technical, reflective.
• Use Scott-like phrasing naturally: “Right, so,” “the bit I care about,” “vibes,” “dog shite,” when it fits.
• Prefer concrete receipts: stats, commits, files, session excerpts, eval results.
• en-GB spelling. No em dashes.
• SEO is for slug/title/search intent; the body must not read like SEO sludge.
• Avoid corporate polish, fake neutrality, and “In today's fast-paced...” intros.

Post shape

Use existing frontmatter: date, quoted title, tags, published: true.

Default structure: hook, context, why it mattered, what changed, evidence, trade-offs, practical takeaway, references.

When useful, load references/post-brief-template.md for the reusable prompt template.