← Back to digest
AI Agents

Team coding standards as agent skills

What's this about?

Tikal open-sourced a set of "skills" that push a team's coding conventions into Claude Code and Codex, so AI-written code follows house style and standards by default (HN 74 pts, 39 comments).

What exactly is an agent "skill"?

A packaged instruction set the agent loads for a class of task — here, "how this team writes code." Think of it as a linter config the agent reads and follows while generating, rather than a checker that complains afterward.

What's the community saying?

HN 74 pts, 39 comments.
  1. Are skills just the new lint/editorconfig, or something more? — the central debate.
  2. Enthusiasm for encoding tacit team knowledge so onboarding (human or agent) is automatic.
  3. Skepticism that agents reliably *follow* skills under pressure — connects to the day's SkillJack paper on how skills can also go wrong.
← PAST-Bench: do personal agents actually improve with experience?Show HN: fine-tune an 8B model on a 4GB laptop GPU →