I found no malicious behavior: the Meta and Kimi code implements provider-specific OAuth login, while the AWS metadata references are in the Bedrock credential path. The observed telemetry is disclosed as an anonymous version/update ping, can be disabled, and reports to pi.dev.
Static reason
High-risk behavior combination matched malicious policy.; previous stored version diff introduced dangerous source
Trigger
Provider OAuth runs during login; update telemetry runs after changelog-detected updates when enabled; AWS metadata resolution is used with the Bedrock provider.
Impact
No harmful impact identified; observed network activity supports the CLI’s provider and update features.
Mechanism
Meta and Kimi exchange login tokens with their respective services. The observed telemetry sends a version and standard user-agent fields to pi.dev. No unrelated credential collection or forwarding was found.
Evidence
package.jsonlib/agent/bundle/agent-entry.jslib/agent/bundle/meta.jslib/agent/bundle/kimi-coding.jslib/agent/bundle/bedrock-converse-stream.jslib/commands/context/platform-context.jslib/upgrade/installation.jslib/upgrade/update-check.js