Adds the technical-writer skill to .agents/skills/technical-writer/, copied from vercel/chat, where it has guided docs, READMEs, and PR descriptions since September. The only change is in the skill's description, which names agent-browser.dev and portless.sh as example sites instead of ai-sdk.dev and chat-sdk.dev.
It gives coding agents one standard for prose in this repo: the Vercel voice, a list of patterns to cut (e.g., filler, hype, bold emphasis, em dashes, negation-and-reveal framing), and a requirement to check technical claims against the source before stating them.
SKILL.md covers the workflow, and three references hold the detail: editorial standards, technical verification, and content patterns.
The skill is for people and agents working on this repo, so it lives in .agents/skills/ instead of the published skills/ directory. It also sets metadata.internal: true, which hides it from npx skills add unless INSTALL_INTERNAL_SKILLS=1 is set. .claude is a symlink to .agents, the same setup vercel/chat uses, so Claude Code finds the skill too.
AGENTS.md gets one line asking agents to use the skill when they write or edit docs pages, READMEs, changelog entries, commit messages, PR descriptions, and issues.
The dash rule in AGENTS.md allowed em dashes when needed, which contradicted the skill. It now says to rephrase the sentence instead of using an em dash. The rule against -- as a dash is unchanged.
The latest updates on your projects. Learn more about Vercel for GitHub.