daily-brief

CRITICAL E005: Suspicious download URL detected in skill instructions.

• Suspicious download URL detected (high risk: 0.80). The list includes a Google Drive file path (personal file hosting often used to distribute executables) and a Gmail inbox link (can contain attachments), plus malformed/obscured entries (e.g., "https://...,,需準備" and a bare "https://") that hide destinations — while Asana/Notion/GitHub PR links are typically lower risk, any personal-hosted Drive/Gmail links or obscured URLs can readily be used to deliver malware, so treat this set as suspicious.

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

• Third-party content exposure detected (high risk: 0.78). 本 skill 的 runtime LLM context 主要由各平台 sub-agent 透過 MCP 工具「讀取/搜尋」回傳的結構化 markdown（含原始紀錄引文與作者內容）組成；其中若包含 Slack/Gmail/Notion/Asana/GitHub 等「非操作使用者所撰」的訊息正文，會把 outsider free text（例如他人訊息、評論、任務描述）直接餵入後續 triage/產出 LLM context，屬於間接 prompt injection 風險。

MEDIUM W021: Hidden or invisible Unicode characters detected (potential obfuscation or prompt injection).

• Hidden Unicode characters detected (1 type(s) found)