The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill utilizes wget to download research papers from user-provided URLs and academic repositories (e.g., Arxiv). It also requires the installation of the notebooklm-mcp-cli tool from a public repository as a backend dependency.
[COMMAND_EXECUTION]: A local shell script (setup.sh) is used for installation, and the agent uses wget for data retrieval. These actions are transparent and consistent with the tool's academic research purpose.
[PROMPT_INJECTION]: The skill has an indirect prompt injection surface as it processes external research papers. Ingestion points: PDF files ingested into NotebookLM (detailed in SKILL.md). Boundary markers: The skill instructs the agent to use grounded queries and exact quotes from the text, creating functional boundaries for the model. Capability inventory: File system writes and network retrieval via wget are used (noted in SKILL.md and setup.sh). Sanitization: The workflow utilizes a grounded extraction model where cited passages are analyzed, reducing the impact of embedded instructions in source data.
[SAFE]: The skill contains no malicious code, obfuscation, or unauthorized data exfiltration. Its design emphasizes local-first storage and researcher supervision of automated processes.

survey