The Agent Skills Directory

[PROMPT_INJECTION]: The skill ingests data from external URLs by fetching the <title> tag to populate database records. While this constitutes an indirect prompt injection surface, it is a primary function of a citation manager and is handled securely using standard libraries.
Ingestion points: Metadata extraction from remote web pages in scripts/source_db.py.
Boundary markers: Retained data is stored in a structured SQLite database and presented in formatted exports (JSON/CSV/Markdown).
Capability inventory: The skill scripts can perform network probes and write to the local file system for database persistence and bibliography exports.
Sanitization: Uses the html.parser library for data extraction and applies standard string normalization to retrieved content.
[COMMAND_EXECUTION]: The skill provides utility scripts for citation management. These scripts are invoked via the agent's shell environment but do not utilize dangerous execution patterns such as os.system with untrusted input, eval(), or dynamic code generation. Database interactions are safely parameterized to prevent SQL injection.

source-tracker