Process files: Blueprints for Agentic AI
summary
Demo shows how a simple markdown playbook turns a general assistant into a repeatable coder: an agent in Windsurf uses the GitHub CLI to fetch an issue, writes a failing test, edits code until tests pass, refactors names and comments, and commits the change for bug #49. The workflow is codified into a .process.md file that tells the agent which tools and commands to run, when to rerun tests (via Maven/run.sh), and when to ask for human confirmation, producing a reproducible, auditable repair loop.
AI-written summary · about
chapters (21)
- 0:00Agentic AI introduction
- 1:03Agentic AI workflow changes
- 8:26Bug fix with AI tools
- 12:35Live demo challenges
- 13:40Choosing a project
- 22:29Project spec with AI
- 26:18Human input in context files
- 28:02Context window management
- 31:58Refining project spec
- 34:57Architecture with AI
- 42:56Process files importance
- 49:10Process file for projects
- 54:30Resuming work with files
- 1:00:19Agentic AI training evolution
- 1:06:45Deterministic vs AI tasks
- 1:15:00Automate build and tests
- 1:21:45Custom files outperform rules
- 1:24:55Main app entry point
- 1:30:15Daily status updates
- 1:37:01Images and emails with files
- 1:39:49Agentic workflows defined
related talks
01h 37mFull Walkthrough: Workflow for AI Coding
41mGoing beyond copilot with AI Agents
29mPairing with the machine: How AI Agents are rewriting the rules of software development
21mBest practices are trapped in your head and your AI has no idea
01h 01m10 tips to level up your ai-assisted coding
40mHow To Tame Your Agents?
32mProgramming language evolution in the AI era
27mAgentic Engineering: Working With AI, Not Just Using It