The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes ssh commands to establish reverse tunnels to the a.pinggy.io service. It utilizes flags such as -o StrictHostKeyChecking=no and -o UserKnownHostsFile=/dev/null to bypass host key verification, which is a common practice for ephemeral automated tunnels but disables standard SSH security checks against man-in-the-middle attacks.
[DYNAMIC_EXECUTION]: The skill provides recipes that generate Python scripts at runtime using shell redirection (e.g., Recipe 1 creating webhook-server.py) and then executes these scripts to handle incoming HTTP traffic. These scripts are generated from static templates provided in the documentation.
[INDIRECT_PROMPT_INJECTION]: The skill describes an ingestion surface for untrusted external data by providing a recipe for a webhook listener that captures and logs request bodies and headers to /tmp/webhook-hits.log. This creates a vulnerability where malicious instructions sent by a third party could influence the agent if it later reads that log file.
Ingestion points: SKILL.md (Recipe 1: /tmp/webhook-hits.log via a local Python server).
Boundary markers: None present in the instructions to separate request data or warn the agent about the untrusted nature of the log content.
Capability inventory: The skill has the capability to run background processes and execute scripts via the terminal tool.
Sanitization: No sanitization or filtering is performed on the incoming request body or headers before they are written to the local log file.
[EXTERNAL_DOWNLOADS]: The skill facilitates networking connections to a.pinggy.io and pinggy.link. Pinggy is a well-known service for developer tunneling infrastructure.

pinggy-tunnel