Skip to main content

​
Intro

Module 3 personalizes the agent stack: capturing reusable instructions and rules, shaping bespoke agents (or borrowing from the library), and wiring MCP tools so every workflow matches your team’s policies and integrations.

​
Video lesson

​
Key takeaways

  • Use global Instructions for always-on guidance, and store reusable Zencoder Rules in .zencoder/rules/*.md so the whole project inherits consistent behaviours.
  • Tweak each rule’s always_apply flag or mention it ad hoc; leverage globs to target file types (e.g., Markdown style guides) and keep shared rules in version control.
  • Remember that repo-info output (repo.md) is itself an auto-applied rule, giving downstream agents immediate architectural context.
  • Build custom agents by setting name, alias, instructions, and tool access; and save them in the custom tab.
  • Browse the agent library for prebuilt roles (architecture reviews, etc.), clone them with one click, and adapt tooling or prompts as needed.
  • Share vetted custom agents with your org by switching scope from personal to org; manage shared catalog entries through the web dashboard.
  • Expand agent capabilities by installing MCP tools: some are one-click (Playwright E2E, Context7 docs) while others require OAuth or API tokens (InstantDB, GitHub).
  • After installation, confirm new tools appear for the right agents and prune unused ones to avoid overwhelming default agent menus.