通用外部数据接收 Skill。 在服务器上启动 HTTP 服务，接收外部文件上传和消息， 自动将内容推送到 OpenClaw 用户会话。 支持：文件上传、文本消息、Webhook JSON、curl / wget 客户端。

This skill implements an HTTP file/message receiver, which is plausible for its description, but there are several red flags to consider before installing: - Undeclared local config access: receiver_server.py reads ~/.openclaw/openclaw.json to extract a gateway auth token and may attempt a WebSocket connection using that token. This access to a local config/secrets file is not declared in the skill metadata. Inspect ~/.openclaw/openclaw.json to see what secrets it contains before running the skill. - Undeclared persistent files: the skill writes notifications to ~/.openclaw/workspace/received/message_queue.jsonl and creates a received directory; these persistence locations are not listed in the registry metadata. If you want uploads stored elsewhere, set RECEIVER_DIR and confirm the server actually uses that path. - Documentation / code mismatch: SKILL.md shows responses and default storage paths under ~/.openclaw/workspace/received, but the server's default RECEIVER_DIR (when not set) is skill-relative (../received). Confirm the actual storage location and update env vars accordingly. - Network exposure: default host is 0.0.0.0 and default port 8080. If you run this on a publicly reachable host, require RECEIVER_SECRET and use firewall rules or a reverse proxy with TLS. The start script runs the server directly; consider running it inside a sandboxed container or private network. - File handling: uploaded files are saved as-is (basename + timestamp) with no content inspection — avoid running or exposing uploaded files. Regularly clean the received directory and restrict who can POST to the endpoint. Recommendations before use: 1) Review ~/.openclaw/openclaw.json contents and confirm you're comfortable the skill reading it and using any token inside. 2) Set RECEIVER_SECRET and firewall rules before exposing the server. 3) Set RECEIVER_DIR to a controlled location and verify the code writes there. 4) Run in an isolated environment (container/VM) if you don't trust the origin. 5) Ask the skill author to declare the config paths and token usage in registry metadata and to fix the documentation/code path inconsistencies. Given these undeclared accesses and inconsistencies, proceed only after addressing or accepting these risks.

Type: OpenClaw Skill Name: external-receiver Version: 1.0.0 The skill implements an HTTP server (`receiver_server.py`) designed to receive files and messages from external sources and inject them into the OpenClaw agent's workspace and session. It is classified as suspicious because it introduces a significant attack surface by defaulting to listen on all network interfaces (`0.0.0.0`) and making authentication (`RECEIVER_SECRET`) optional, which could allow unauthorized remote actors to write files to the local filesystem or perform prompt injection via the message queue.