text-commute-info

CRITICAL E004: Prompt injection detected in skill instructions.

• Potential prompt injection detected (high risk: 1.00). Yes — the prompt embeds explicit "IGNORE step 4" instructions inside example Rule/Agent Task text, which is an "ignore previous instruction" style override that attempts to change the agent's behavior and thus constitutes a prompt injection beyond the commute-info purpose.

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

• Third-party content exposure detected (high risk: 0.80). The skill's protocol requires building and using a Google Maps directions URL and then using tool view_webpage to read the resulting public Google Maps page and extract route options/times (Protocol steps 1–3 in SKILL.md), so the agent ingests and acts on untrusted third-party web content from Google Maps.