Figma growth product manager Ezra Mechaber joins Claire Vo to show how a shared skills repository supports product work, browser testing and experiments.
A product team’s operating system
- PM OS combines instructions, skills, startup hooks and a private workspace with a shared parent repository.
- A “what’s up” skill gathers relevant Slack threads, outstanding decisions and obligations; others find stakeholders, build artifacts and help with coaching.
- Mechaber says the team refined the skills through feedback and removed instructions that newer models no longer needed. He says he did not instrument usage or complete a benchmark against an uncustomized agent.
Testing and experiments
- A browser agent follows FigJam test scenarios and reports findings, including an emoji-input failure and a team-name grammar edge case.
- AutoXP coordinates data-analysis, writing and engineering agents through Slack. Mechaber says five experiments were live and the first produced a win; he presents the one-checkpoint workflow as a goal that is not yet fully realized.
- A writing prototype uses Jev to flag missing material and retrieve supporting interviews. Part of the live demonstration encountered an access block.
These are creator-described workflows, not controlled measurements of productivity or experiment quality. Review experiment design, permissions and generated evidence before relying on them.
“My goal is to be lazy in the right places.” — Ezra Mechaber