tunneling

Create free SSH tunnels to expose local ports to the internet using tinyfi.sh. Use when you need to share a locally running app, test webhooks, demo a prototype, or get a public HTTPS URL for any local service — no signup or authentication required.

Before installing or letting an agent run this skill, consider: (1) The metadata fails to list 'ssh' as a required binary even though the instructions rely on it — ask the publisher to correct that. (2) tinyfi.sh has no listed homepage or source in the skill; verify who operates tinyfi.sh and whether you trust that operator, because all tunneled traffic (and URLs) will pass through their servers. (3) The recommended ssh option StrictHostKeyChecking=accept-new auto-accepts host keys — this can allow man-in-the-middle connections; prefer verifying the server fingerprint manually. (4) Never expose services that handle secrets, authentication tokens, private keys, or production databases via an untrusted tunnel. If you must test webhooks/demos, run on ephemeral test data behind authentication. (5) If you don't fully trust tinyfi.sh or autonomous agent execution, run the ssh command yourself manually and inspect the connection, or use a well-known tunneling provider (or your own SSH server) with documented security practices. (6) Consider limiting the agent's ability to execute background network processes unless you explicitly trust the skill and its operator.

Type: OpenClaw Skill Name: tunneling Version: 1.0.0 The skill is designed to create SSH tunnels to expose local ports using the tinyfi.sh service. It instructs the agent via SKILL.md to execute an `ssh` command in the background to `tinyfi.sh`, which is directly aligned with its stated purpose. While `ssh` is a powerful tool, its use here is for its intended tunneling functionality, and there is no evidence of data exfiltration, malicious execution beyond the stated purpose, or prompt injection with harmful objectives. The `StrictHostKeyChecking=accept-new` flag is a minor security convenience but does not indicate malicious intent.