telnyx-voice-java

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (medium risk: 0.30). Telnyx Voice webhook handler path ingests Telnyx-delivered HTTP POST payloads via client.webhooks().unwrap(...body(payload)... ), where webhook JSON fields may contain outsider-authored free text (e.g., data.payload.custom_headers / sip_headers / from/to) before the agent processes the event.

HIGH W008: Secret detected in skill content (API keys, tokens, passwords).

• Secret detected (high risk: 1.00). I flagged the high-entropy "v3:..." strings because they look like opaque call-control tokens (the docs describe callControlId as "Unique identifier and token for controlling the call") — these are literal, high-entropy values that could be used to control calls, so they meet the definition of a secret.

Ignored items / why not flagged:

• "7267xxxxxxxxxxxxxx" is a masked/placeholder connectionId (low entropy / intentionally redacted).
• Phone numbers such as "+18005550101" and numeric IDs like "1293384261075731461" are not secrets.
• UUIDs used as examples (e.g., "550e8400-e29b-41d4-a716-446655440000") are common example values and by themselves are not high-entropy access tokens unless combined in a token (in this case they appear as part of the v3 token which I did flag).