The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: Recommends the pi-extension-config library for type-safe configuration management, which is an official resource from the vendor.
[COMMAND_EXECUTION]: Provides implementation patterns for the pi.exec function, enabling extensions to interact with external tools like git through controlled subprocess calls.
[PROMPT_INJECTION]: The skill documentation describes the ingestion of untrusted data from user inputs that could influence agent behavior.
Ingestion points: Extension commands receive input via arguments and interactive TUI elements such as text fields.
Boundary markers: Patterns shown do not include specific delimiters or instructions to prevent command injection in processed data.
Capability inventory: Examples demonstrate the use of pi.exec and node:fs for core functionality.
Sanitization: Snippets include validation logic such as regular expression checks for input strings and viewport clamping for UI elements.

creating-pi-extensions