The Agent Skills Directory

[COMMAND_EXECUTION]: The skill invokes several macOS system utilities using execFileSync, specifically sqlite3, zip, unzip, open, and osascript (AppleScript).
SQL queries are executed with the -readonly flag to ensure the live database is not modified.
The skill implements custom sanitization functions, such as q() for escaping SQL string literals and safeName() for cleaning filesystem paths, to mitigate command injection risks.
The use of osascript for GUI automation is documented and requires the user to grant macOS 'Automation' permissions to the terminal.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and structure untrusted text from external sources such as web pages, PDFs, and notes into the Paprika recipe format.
Ingestion points: The agent is instructed in SKILL.md to read content from various external sources (e.g., PDF text, note contents, web page text) to generate JSON recipe objects.
Boundary markers: The skill does not implement specific boundary markers or instructions to the agent to disregard embedded commands within the ingested source text.
Capability inventory: The skill scripts have the capability to write files, execute shell commands, and drive the macOS GUI through AppleScript.
Sanitization: While the skill sanitizes inputs used in shell commands and database queries, it does not sanitize the text content being processed by the agent into the recipe fields.

paprika-recipes