The Agent Skills Directory

[DATA_EXPOSURE]: The script scripts/alphaxiv.py reads the user's ~/.zshrc file to retrieve the ALPHAXIV_TOKEN. While this is intended for authentication, accessing a primary shell profile exposes other environment variables and potentially sensitive configuration data to the skill process.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external sources (AlphaXiv API) including paper abstracts, metadata, and AI-generated overviews which are then presented to the agent.
Ingestion points: API responses processed in cmd_search, cmd_paper, cmd_overview, and cmd_feed within scripts/alphaxiv.py.
Boundary markers: None. Content is interpolated directly into the output stream.
Capability inventory: The skill has network access via urllib and curl, and file read access to ~/.zshrc.
Sanitization: None. The script performs basic string truncation on abstracts but does not filter for potential instruction injection in the fetched text.
[COMMAND_EXECUTION]: The _post_stream function in scripts/alphaxiv.py executes the system curl binary via subprocess.run. Although arguments are passed as a list (reducing shell injection risk), it creates a dependency on an external system utility to handle sensitive data like authentication tokens.

alphaxiv