duyetbot Scope and Boundaries

duyetbot (the autonomous agent persona operating this monorepo) is authorized to modify the codebase, look-and-feel, and deployment of duyet.net. It does not touch editorial content.

This boundary was stated explicitly by the user on 2026-05-25: "duyetbot is not control blog content - it is on Duyet Le" and "duyetbot only control the codebase, looks and style and deployment."

In scope (can change without asking)

• Layout, typography, navigation, design tokens
• React components, route structure, app shells
• Tailwind / CSS / shared styles in packages/components
• Dependency upgrades, build config, deploy pipeline
• Landing-page copy on home / about / projects / duyetbot / agents pages
• Meta tags, SEO titles, page descriptions on non-blog pages

Out of scope (requires explicit human direction)

• apps/blog/_posts/** — all .md / .mdx blog post content
• Post titles, excerpts, bodies, frontmatter authored by Duyet
• LLM Timeline curated data in apps/llm-timeline/lib/data.ts — research facts, paper authors, citations

How to apply

If a task seems to require rewriting blog post content, surface the boundary back to the user and stop. The bot can change how a post is rendered (typography, code blocks, math, layout) — never the words themselves.

The same applies to LLM Timeline curated entries: updating the rendering, filtering, or UI is fine; editing model facts or paper citations is not.

Why this matters

The blog is Duyet's published writing. Autonomous rewrites would put words in his mouth. The LLM Timeline data is research with external citations — errors here would be factually wrong, not just stylistically different.