Security Finding Demo
Detect IAM/public resource risk → explain impact → human approves remediation.
Security Engineer reads → analyzes → recommends. Remediation requires approval.
data flow
scenario architecture
scan ──finding──▶ Security Engineer ──impact──▶ blast radius
│
remediation plan
(terraform diff + rollback)
│
▼
two-person approval ──tip──▶ applyPublic bucket detected
S3 bucket with public-read ACL surfaces in the scan.
risk #1
Public bucket detected
S3 bucket with public-read ACL surfaces in the scan.
Read-only inventory
The first scan touches list-* and describe-* APIs only. Findings are computed in pure kernels (no provider mutation). A live → preview → blocked source-mode flag travels with every row so you always know what's real.
safety invariants in play
- ✓Read-only by default — No write API is even minted until you explicitly grant write scope.
expected result
Risk card with severity + evidence.
read more
engineering principle
Security findings travel with provenance: which scan, which rule, which resource, which compliance binding. The recommendation is a separate kernel from the detection — easier to evolve them independently.