adk-deploy-guide

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.90). The skill includes concrete shell examples that embed secret values (echoing API keys into gcloud and passing secrets as make/CLI arguments), which encourages or requires the LLM to emit secrets verbatim in commands or deployment strings.

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

• Third-party content exposure detected (medium risk: 0.30). Skill runtime path /apps/my_agent/trigger/pubsub, /apps/my_agent/trigger/eventarc, and the BigQuery remote function handler / decode and pass outsider-provided free text from push/event payloads (body/message/data base64-decoded or data/calls JSON) directly into _run_agent(..., new_message=types.Part(text=message_text)) for LLM processing.