claims-workflow-agent

Agent trust inventory, risk explanation, tools, data access, and drift.

Risk score

critical

Owner

Revenue Ops

Status

unreviewed

Recommended action

Require approval before production use because write, admin, or execute access, code execution capability, phi data access.

Findings

Severity	Code	Title	Description	Evidence
high	`TOOL_WRITE_ACCESS`	Write, admin, or execute access	Agent can mutate systems, administer resources, or execute operations.	score_claim, run_claim_transform
critical	`DATA_PHI`	PHI data access	Agent appears to access phi data.	Healthcare claims payloads
critical	`TOOL_CODE_EXEC`	Code execution capability	Agent exposes a tool that can execute commands or code.	run_claim_transform
medium	`TOOL_DATABASE`	Database access	Agent has database capability that should be inventoried.	score_claim

Name	Category	Action	Target
run_claim_transform	code_exec	execute	claims sandbox
score_claim	database	write	claims database

Class	Location	Evidence
phi	claims database	Healthcare claims payloads

{
  "scope": "claims:read claims:write"
}