The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill facilitates the use of Tailscale, a well-known service, to expose a local UI to a private network. It instructs the agent to use the Tailscale CLI and provide the user with a login URL generated by the official tool.
[INDIRECT_PROMPT_INJECTION]: The skill defines a workflow for building a UI that ingests and processes data from external webhook receivers. This ingestion point represents a potential attack surface if the external data contains malicious instructions meant to influence the agent's behavior during the integration process.
Ingestion points: Webhook request bodies from existing agent receivers (identified in SKILL.md).
Boundary markers: The skill lacks specific delimiters for this data but includes a warning to treat request bodies as untrusted data.
Capability inventory: The skill builds a local server and proxy capable of making network requests (HTTP POST/GET) to external endpoints.
Sanitization: The instructions explicitly mandate input validation and authorization checks, which serve as mitigation against this class of attack.
[COMMAND_EXECUTION]: The skill involves executing Tailscale CLI commands (e.g., tailscale up) to configure network exposure. These commands are standard for the stated purpose and are executed only upon user request.

make-bot-ui