Bỏ qua tới nội dung
TakoAPI
Quay lại kỹ năng

mirage-proxy

bởi chandika

Web & Frontend Development
381Lượt 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

Install and configure mirage-proxy as a transparent PII/secrets filter for OpenClaw LLM API calls.

Mô tả

Kỹ năng này làm gì

mirage-proxy sits between OpenClaw and LLM providers, intercepting API calls to redact secrets and PII before they reach the model. It replaces sensitive values with plausible fakes so the LLM processes realistic-looking but sanitized data. A bundled setup script handles binary download, proxy startup, and provider config patching.

Unlike per-agent prompt sanitization, mirage-proxy intercepts all LLM traffic at the network layer without requiring changes to any agent code.

Khi nào nên dùng

  • Reviewing codebases that contain hardcoded API keys or credentials
  • Feeding customer records to an LLM without exposing real names or emails
  • Debugging agent prompts that pull values from environment variables
  • Comparing filtered vs unfiltered model responses side-by-side using model aliases
  • Running compliance-sensitive workflows where secrets must not leave the system unredacted

Quy trình ví dụ

ĐẦU VÀO

User asks an agent to review a .env file containing AWS access keys

AGENT
  1. 1

    Agent sends file contents in a prompt routed through the mirage-anthropic provider

  2. 2

    Proxy intercepts the outbound request and detects the AWS Access Key pattern

  3. 3

    Proxy replaces the real key with a plausible fake before forwarding to Anthropic

  4. 4

    LLM receives and processes the sanitized prompt

  5. 5

    Agent returns the response; real credentials were never transmitted

ĐẦU RA

LLM analysis of the config file with no real secrets exposed in transit

Cài đặt

clawhub install chandika/mirage-proxy

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

Liên kết