meticulous-use-session-data

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The skill tells the agent to read full request/response bodies and storage (cookies, localStorage, sessionStorage) and to "use the response.content.text" as mock data in tests, which requires embedding verbatim data that may include API keys, tokens, or cookies — creating an exfiltration risk.

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

• Third-party content exposure detected (high risk: 0.70). The SKILL.md requires running "meticulous local relevant-sessions" to download recorded session data and explicitly instructs the agent to read network-requests/.json (including response.content.text) and user-events.json — i.e., ingesting user-generated/untrusted third-party session and network content — and to use that data as mocks and to guide testing, so external content is read and can materially influence subsequent actions.