dogfood

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The skill explicitly instructs the agent to insert user credentials into shell commands (e.g., agent-browser fill "{EMAIL}" / "{PASSWORD}") which requires the LLM to output secret values verbatim, creating an exfiltration risk.

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

• Third-party content exposure detected (high risk: 0.70). Required workflow uses agent-browser ... open {TARGET_URL} and then takes snapshots/screenshots and reads console/errors from the target web app; any outsider-authored page content/console text could be ingested into the agent’s LLM context via the browser tooling outputs.