The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to its core function of ingesting and analyzing external data from research papers.
Ingestion points: The skill reads external content from arXiv (via HTML or abstract pages) and fetches metadata from the Semantic Scholar API (SKILL.md).
Boundary markers: There are no instructions to use delimiters or ignore potential embedded instructions within the papers, creating a path for attackers to influence agent behavior through malicious paper content.
Capability inventory: The agent possesses capabilities to execute shell commands via the notebooklm CLI and perform web requests, which could be misused if an injection is successful.
Sanitization: The skill does not define any validation or filtering steps for the data retrieved from external academic repositories.
[EXTERNAL_DOWNLOADS]: The skill retrieves research data from reputable external repositories.
Evidence: Fetches paper metadata from api.semanticscholar.org and full content from arxiv.org (SKILL.md).
[COMMAND_EXECUTION]: The skill instructs the agent to utilize a command-line tool for managing research data.
Evidence: Orchestrates several commands using the notebooklm CLI, including notebooklm use, notebooklm source add, and notebooklm ask (SKILL.md).

paper-analyzer