Best Practices
Smart habits in Weaver will save you late-night bug-hunts and surprise token bills.
TL;DR
Why Best Practices Matter
Hours saved debugging ≠ hours you can spend writing release notes about how little debugging you did.
Tokens = money; broken flows = sad work DMs. Let’s stay happy.
How to Build, Test, and Deploy Your AI Agent Workflow
Build small -> Test early -> Deploy to staging -> Ship to production.
Build Small
Add one skill at a time. If it fails, you know the culprit instantly.
Test Early
Toggle Debug Mode every run until borders turn 🟢 green.
Deploy Staging
Push to a dev subdomain first; let the intern break it before customers do.
Embrace Modularity
- Break big goals into numbered skills (Weaver parses lists).
- Use existing Templates (see Agent Templates) or remix them.
- Keep each skill single-purpose.
Naming Conventions
Item | Convention | Example |
---|---|---|
Agent | Verb + Outcome | Generate SEO Article |
Skill | Action + Object | Extract Keywords |
Inputs | snake_case | max_results |
Performance Baseline
Follow these simple tips
Optimise | How |
---|---|
Choose models wisely | GPT-4o mini beats full GPT-4o on price |
Throttle loops | Drop Sleep Component after 10 iterations |