cdp

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 0.90). The script intentionally executes arbitrary JavaScript supplied via the --measure flag inside the target page (page.evaluate) and includes returned values in its JSON output, which can be trivially abused to read and exfiltrate sensitive page data (cookies, localStorage, tokens) or to run harmful payloads against visited pages.

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

• Third-party content exposure detected (medium risk: 0.30). The runtime path tools/cdp.ts ingests outsider-authored free text via the --shot argument (label|url) and then loads that URL at page.goto(shot.url, ...), running attacker-controlled JavaScript in page.evaluate(options.measure) and capturing DOM-dependent rendering.