Skill Issue: How We Used AI to Make Agents Actually Good at Supabase
summary
Workshop claims that concrete Skills + evals improve agent behavior and demonstrates it: agents use MCP and CLI tools, Claude-powered sessions, and a Braintrust eval harness to build a Supabase demo view, discover missing row-level security, and add the security_invoker flag so RLS applies. The demo app and scripts exist, evaluations run headless, results are mixed—sometimes the Skill changes behavior and adds the flag, sometimes the model output or grading misses it—showing progressive disclosure and measurable iteration work.
AI-written summary · about
chapters (12)
- 0:00Introduction to skills
- 3:06Defining agent skills
- 4:04Skills vs MCP tools
- 8:13Testing with evaluations
- 11:33The evaluation lifecycle
- 15:21Demo application setup
- 19:42Implementation without skills
- 26:17Identifying security gaps
- 32:45Writing the security skill
- 37:53Installing and using skills
- 44:24Refining agent behavior
- 52:59Advanced evals and conclusion
related talks
29mStop Prompting, Start Composing: Agent Skills in Practice
31mSpecialize Your Agents with Skills
27mSkills are the New SDKs
15mEnsure AI Agents Work: Evaluation Frameworks for Scaling Success
07h 47mAI Agents Conference 2026 - Day 5
01mSpecialize Your Agents with Skills
30mBuilding Reliable AI Agents: An Open Source Approach To Evaluation and Observability
08mBuilding an AI Platform Your Engineers Actually Trust