🔒AWS Open-Sources Dogwood: Policy Language for Agent Tool Calls
Dogwood adds temporal conditions to agent policies
TL;DR
AWS has open-sourced Dogwood, a policy language that extends Cedar's capabilities by adding temporal conditions. This allows for more complex rule sets based on historical events and actions.
AWS just dropped Dogwood, a new policy language built to enhance agent tool calls with temporal conditions. If you're managing complex workflows or security policies in your AWS environment, this is a big deal because it lets you write rules that look back at what the system did before. For instance, if an action was taken 5 minutes ago and needs to be approved based on past events, Dogwood makes that possible with its four operators: formerly, count_within, count_distinct_within, and sum_within. This is a game changer for teams dealing with intricate security or operational policies where the sequence of actions matters.
Key Points
Dogwood's rules can look back at what the agent already did, using four operators: formerly, count_within, count_distinct_within, and sum_within.
The language is open-sourced under Apache 2.0 but isn't ready for production use yet; it's meant for testing and exploration.
AgentCore Policy supports Dogwood today, expanding on Cedar's capabilities to include historical context in policy enforcement.
Temporal conditions require stateful tracking of events, meaning durable storage solutions are essential for implementing Dogwood effectively.
Dogwood's roadmap includes absolute-time windows and liveness properties, promising even more sophisticated rule sets in the future.
Why It Matters
If you're managing complex security policies or workflows with AgentCore Policy, Dogwood introduces a way to enforce rules based on historical context. For instance, teams can now write policies that consider past actions when deciding whether to approve current requests, making it easier to handle intricate scenarios without manual oversight.
Frequently Asked Questions
Why does this matter?
If you're managing complex security policies or workflows with AgentCore Policy, Dogwood introduces a way to enforce rules based on historical context. For instance, teams can now write policies that consider past actions when deciding whether to approve current requests, making it easier to handle intricate scenarios without manual oversight.
What happened?
AWS has open-sourced Dogwood, a policy language that extends Cedar's capabilities by adding temporal conditions. This allows for more complex rule sets based on historical events and actions.
Comments
Be the first to comment
Enjoyed this article?
Get it daily. 7am. Free. Reads in 5 minutes.
Join 3,049 builders reading daily.