The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions mandate the execution of the open <filename>.ics command in the shell after writing the file. This behavior uses a system-level utility to launch applications based on file extensions.
Evidence: The Workflow and Defaults sections in SKILL.md explicitly instruct the agent to run the open command.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted natural language input from the user to generate structured file content and filenames, creating a potential surface for injection attacks.
Ingestion points: User-provided natural language event descriptions (summary, location, description, etc.) parsed by the agent.
Boundary markers: None explicitly defined for the input stage; however, the skill defines a strict VCALENDAR schema for output.
Capability inventory: The skill writes files to the current working directory and executes the open shell command.
Sanitization: The skill includes explicit instructions for 'Text Escaping' of backslashes, semicolons, commas, and newlines within the file body, and specifies 'kebab-case' for filenames, which serves as a basic sanitization layer for the command-line argument.

ics-generator