Bỏ qua tới nội dung
TakoAPI
Quay lại kỹ năng

agentcloak-email-proxy

bởi ryanfren

Coding Agents & IDEs
391Lượt tải do clawskills.sh công bố
0Lượt xem
0 lượt thíchĐăng nhập để thích kỹ năng này
—Chưa có đánh giá
Đánh giá của bạn

Tóm tắt

Secure email proxy for AI agents.

Mô tả

Kỹ năng này làm gì

AgentCloak sits between an AI agent and your email account, providing search, read, and draft access through a 4-stage security pipeline (blocklist, HTML sanitization, PII redaction, prompt injection detection). Credentials stay server-side; the agent only sees filtered content. Available as a hosted service or self-hosted instance.

Unlike raw email skills that give agents unfiltered inbox access, AgentCloak runs credentials and content through a server-side security pipeline so the agent never sees passwords, PII, or prompt injection attempts.

Khi nào nên dùng

  • Finding unread emails without exposing email credentials to the agent
  • Drafting reply emails for human review before sending
  • Reading full conversation threads for context while filtering PII
  • Blocking financial and security-sensitive emails from agent access
  • Checking label-based unread counts across the inbox

Quy trình ví dụ

ĐẦU VÀO

User asks: Find my unread emails from this week and draft a reply to the most recent one from my manager

AGENT
  1. 1

    Calls search_emails with query 'is:unread newer_than:7d' to retrieve recent unread messages

  2. 2

    Calls read_email on the most recent result to get filtered, PII-redacted content

  3. 3

    Identifies the thread and sender from the sanitized email body

  4. 4

    Calls create_draft with a composed reply and in_reply_to_thread_id set to the original thread

ĐẦU RA

Draft reply created in the email account, ready for user review before sending

Yêu cầu

  • AGENTCLOAK_API_KEY environment variable
  • AgentCloak account at agentcloak.up.railway.app (for hosted version) or self-hosted instance with email connected

Cài đặt

clawhub install ryanfren/agentcloak-email-proxy

Chạy lệnh CLI, hoặc dán prompt vào Claude Code, Cursor, Codex hay coding agent bạn đang dùng.

Liên kết