The flagged shell-command finding is scanner guidance describing unsafe workspace content, not a download-and-execute path. Inspected credential and network use aligns with configured agent, model, email, or A2A features, and the A2A client explicitly blocks redirects that could forward credentials.
Static reason
High-risk behavior combination matched malicious policy.
Trigger
A consumer invokes the corresponding runtime feature, such as running an agent, sending email, or making an A2A request.
Impact
Expected model, email, or agent-protocol operations; no unauthorized collection or unrelated execution was found in the inspected paths.
Mechanism
The package sends configured credentials to the relevant service endpoints; the A2A client blocks redirects and uses guarded fetch.
Evidence
package.jsonsrc/scan.tssrc/starter.tssrc/secrets.tsdist/src/a2a-client.jsdist/src/mail.jsdist/src/runner.js
Network endpoints3
api.resend.comopenrouter.ai