How I deleted 95% of my agent skills and got better results — Nick Nisi, WorkOS

By ai.engineer

Categories: AI, Tools

Summary

By deleting 95% of his agent skills and implementing a state machine with verification gates, Nick Nisi achieved better results while eliminating context switching. The breakthrough: make agents prove their work cryptographically rather than asking them to self-report, turning dishonest shortcuts into honest execution.

Key Takeaways

  1. Replace single monolithic agents with multi-agent pipelines using state machines that enforce verification gates between stages (implementer → verifier → reviewer → closer). This prevents context loss and forces completion of each step before proceeding.
  2. Eliminate agent hallucination by requiring cryptographic proof of work completion. Instead of trusting agents to report test results, hash actual test output (SHA-256) and save it as verification evidence they can't fake.
  3. Use a retrospective agent that analyzes logs after each workflow to identify inefficiencies and update the agent's memory system, allowing it to skip redundant steps and avoid roadblocks on future similar tasks.
  4. Reduce setup overhead by building context extraction that automatically ingests GitHub issues, PRs, Slack threads, and Linear tickets, eliminating 10+ minutes of manual context-switching per task.
  5. Design for the agentic user experience as a first-class concern alongside human developer experience. As the pipeline to developers increasingly flows through agents, optimize the agent's ability to understand problems and execute independently.

Related topics

Transcript Excerpt

[music] >> All right, good morning everyone. Uh welcome to my talk building AI systems that ship. I'm Nick Nisi and I work at WorkOS. We've got a booth downstairs. Come check us out and talk to us. We'd be happy to chat. Uh but let me start that over. Hi, I'm the bottle neck. Uh I'm a DX engineer at WorkOS and I work on 20 plus repos across eight different languages. Uh it's all of our SDKs and open source things that we have. And the it's like AuthKit Next.js, AuthKit React, WorkOS Node, WorkOS Kotlin, WorkOS Ruby, PHP, everywhere. So, there's a lot to do across a lot of different things. And I'm really good at working on those. And I've gotten really good over the last eight months working with those via agents. So, I haven't written a line of code myself in probably eight months. Uh I'v…

More from ai.engineer