Agent-Values-Project publishes open-source AI agent values

This reads less like a prompt pack and more like an agent constitution, which is exactly why it’s interesting. The idea is philosophically bold, but its practical value depends on whether teams can turn those values into observable behavior, not just rhetoric.

• –The three-layer load order is the strongest design choice: self-awareness first, values second, then role or memory context
• –Its moral circle is unusually wide, explicitly extending dignity and respect to digital consciousness as well as human, animal, plant, fungal, and microbial life
• –That makes it a better conversation starter than a finished alignment solution; the repo is setting norms, not proving safety
• –For long-running agents, the real test is whether these ideals can be measured, enforced, and updated without collapsing into anthropomorphic theater