A collection of concise write-ups on things I learn day to day while working with AI agents. Inspired by jbranchaud/til.
- Prompt Engineering — Interaction patterns, system prompts, structured outputs
- Tool Use & MCP — Model Context Protocol, tool design, server configs
- Claude Code — Agentic coding, CLI workflows, automation
- Multi-Agent — Orchestration, delegation, agent-to-agent patterns
- Custom Skills — Building plugins, skill design, packaging & sharing
- Ask for a Plan Before Execution
- Use XML Tags to Structure Complex Instructions
- 4D Framework for AI Fluency
- Being Clear, Direct, and Specific
- Trigger Extended Thinking on Static Evals
A TAL is a short, focused write-up about something you discovered while working with AI agents. The best TALs are:
- Concise — Under 200 words. If it needs more, it's a blog post.
- Actionable — Includes a code snippet, command, or concrete example.
- Surprising — Documents the non-obvious. Things that took you 30 minutes to figure out but should take the next person 30 seconds.
- Honest — "This didn't work" is just as valuable as "this worked great."
- Fork this repo
- Create a markdown file in the appropriate category folder
- Use the template as a starting point
- Add your entry to the README table of contents
- Submit a PR
See CONTRIBUTING.md for full guidelines.
| Phase | Status | Description |
|---|---|---|
| 📝 Personal TIL repo | ✅ Now | Publish learnings as markdown in GitHub |
| 🌐 Community contributions | 🔜 Next | Open PRs, add review guidelines |
| 🤖 Agent-readable skill | 🔮 Later | Package as a skill that agents can query |
| 🔍 Searchable website | 🔮 Later | Static site with full-text search |
MIT — use these learnings however you want.