JFrog has documented an on-behalf-of exchange so agent calls reach Artifactory as the signed-in user. The price is the Gateway's cached tool search, which per-user discovery gives up.
Reality
- Evidence55
- Adoption
- Not scored
- Hype gap+15
- Incentives70
- Confidence60
buildOne report1 publisher Sharing long-term memory across two agents comes down to the actorId you write under
A dev.to walkthrough gives a Nova Sonic voice agent and an Amplify AI Kit chat agent one long-term memory in Amazon Bedrock AgentCore Memory, where every record is filed under an actorId and a sessionId. Both channels already had short-term context.
Reality
- Evidence58
- Adoption12
- Hype gap+14
- Incentives52
- Confidence58
buildOne report1 publisher The default MCP deployment buys a Kubernetes operational surface for request/response traffic
A dev.to writeup makes the case for trying Bedrock Agents or AgentCore Runtime before containers, and for putting real limits on the gateway, because a looping agent spends money without ever raising an error.
Reality
- Evidence28
- Adoption
- Not scored
- Hype gap+18
- Incentives32
- Confidence36
buildOne report1 publisher Nx Plugin for AWS 1.0 generates four KMS keys and three web ACLs before the first feature lands
One developer took Nx Plugin for AWS 1.0, the AWS Blocks preview and Amplify Gen 2 to the same file-portal scope, deployed all three against one FSx for ONTAP access point, then deleted everything to see what stayed behind.
Reality
- Evidence58
- Adoption25
- Hype gap−5
- Incentives30
- Confidence55
buildOne report1 publisher Cognito's replica user pool rejects every attribute write with OperationNotEnabledException
AWS's multi-Region replication copies a Cognito pool's configuration, hashed credentials and IdP settings into one other Region with eventual consistency. Only the primary accepts writes, so failover planning becomes a per-path decision.
Reality
- Evidence66
- Adoption
- Not scored
- Hype gap+22
- Incentives85
- Confidence64
buildOne report1 publisher A leftover localhost callback lets Cognito hand OAuth tokens to whatever binds port 3000
Cognito validates the redirect target by matching it against the app client's allowed callback list, so a development entry nobody removed is a valid destination for an authorization code, or for the tokens themselves where implicit grant is still on.
Reality
- Evidence45
- Adoption
- Not scored
- Hype gap+35
- Incentives35
- Confidence50
buildOne report1 publisher Cognito ships a new user pool with MFA off and its sign-in risk model disabled
A pool created with the defaults treats a password as the whole authentication factor and scores no risk on the sign-in. In the HackerOne chain a dev.to writeup walks through, either setting turned on would have stopped the takeover.
Reality
- Evidence55
- Adoption
- Not scored
- Hype gap+15
- Incentives68
- Confidence52
buildOne report1 publisher Why a 14-day Google Play test window pushed a 19-language retrofit before launch
A survey found roughly 660 user-facing text nodes across 53 files and no i18n at all. What decides whether thirteen unreviewed locales survive is a plural-rules polyfill and the choice to store locale per user.
Reality
- Evidence45
- Adoption10
- Hype gap+20
- Incentives55
- Confidence40
buildOne report1 publisher Bedrock's retrieval planner decides how many hops must carry the tenant filter
Agentic retrieval turns one question into an unknown number of lookups, and each is a place isolation can fail, which is why the interesting line in AWS's design is the one still leaving per-user isolation with your application.
Reality
- Evidence34
- Adoption
- Not scored
- Hype gap+26
- Incentives88
- Confidence48
One unvalidated region string sent signed AWS API calls to attacker.com
AWS patched seven SDKs and called it a defense-in-depth enhancement. Pi says all seven third-party deployments it tested handed over live credentials, which puts the region field on your audit list.
Reality
- Evidence58
- Adoption62
- Hype gap+16
- Incentives74
- Confidence55
buildOne report1 publisher AWS's one-minute test for agent access is really a test of where the answer lives
Amazon says you should be able to name every AI agent touching customer data in under a minute. Its own worked example explains why most teams cannot: the record is a file on a laptop.
Reality
- Evidence34
- Adoption
- Not scored
- Hype gap+28
- Incentives82
- Confidence46
buildOne report1 publisher Cognito rejects RFC 8707, so it cannot be the authorization server for an MCP client
A dev.to writeup documents Amazon Cognito killing a spec-compliant MCP connector at the first redirect, and the only fix on offer is a proxy that strips one parameter and nothing else.
Reality
- Evidence46
- Adoption18
- Hype gap−6
- Incentives30
- Confidence44
buildOne report1 publisher AWS moves agent authorization out of the agent and into the plumbing
A new Bedrock AgentCore pattern from AWS treats the agent as an orchestrator with no say over access, pushing per-user enforcement down to DynamoDB, Knowledge Bases and Salesforce.
Reality
- Evidence56
- Adoption
- Not scored
- Hype gap+22
- Incentives80
- Confidence58
buildOne report1 publisher Axonius runs one agent per tenant on AgentCore, and tracks model cost the same way
AWS published the build pattern most agent demos skip: isolation and cost attribution scoped per customer, not per application. Axonius chose silo and kept its existing tenant plumbing.
Reality
- Evidence42
- Adoption28
- Hype gap+12
- Incentives85
- Confidence48