Poteto mode: verification is the infrastructure

The playbook starts from trust. In her talk "here's how i shipped 2,500 PRs last month to production", lauren says the reason people get stuck babysitting one to five agents is that they do not trust the work yet. Her first fix was verification: a skill that teaches the agent to run the app, take traces, and collect evidence that the code works. Her team found these skills so useful that they became more or less critical infrastructure, and they maintain them constantly.

Verification covers correctness, not quality. For quality she built pstack, a plugin of skills drawn from her own engineering workflows, such as debugging, feature development, and prototyping. Its README sets the bar: "if you want to go fast, go deep first." Its poteto-mode skill describes the style as concise, detailed responses, deliberate subagents, unslopped prose, simple code, and verified work, guided by 24 principles that include Laziness Protocol, Subtract Before You Add, Experience First, Prove It Works, and Model the Domain. Under the skills sit harder constraints: a codebase where the mistake is categorically impossible, then lint rules, then rules and code review bots.

A template is listed here only when its own name, skills, or memories follow poteto mode as a coding loop: work handed to cloud agents, a verification skill as the bar for done, no local writing. Bots that delegate coding differently, such as steering one long-lived agent, are not listed.