agent-browser

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The prompt's examples and workflows explicitly show filling forms and running commands with literal passwords and environment-variable expansions (e.g., agent-browser fill @e2 "password123" and agent-browser fill @e1 "$USERNAME" / fill @e2 "$PASSWORD"), which requires the agent to include secret values verbatim in generated commands or outputs, creating an exfiltration risk.

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

• Third-party content exposure detected (high risk: 1.00). The skill explicitly navigates to and ingests arbitrary external web pages (e.g., SKILL.md's "agent-browser open " workflow and templates/capture-workflow.sh which open and snapshot arbitrary URLs) so untrusted third‑party page content is read and used to drive interactions and decisions (snapshots, get text, click, eval), enabling indirect prompt injection.