The Agent Skills Directory

[SAFE]: The skill prioritizes security by explicitly forbidding automated login attempts, requiring users to authenticate manually and halting if a login wall is detected.\n- [SAFE]: It employs a robust browser automation discipline, mandating screenshots before every action and using semantic element finding to maintain stability against UI changes.\n- [SAFE]: The skill identifies and mitigates the risk of Indirect Prompt Injection by structuring data ingestion workflows and recommending pre-processing of external content. Ingestion points: 'Add Source' sub-flows in SKILL.md; Boundary markers: Clean formatting of chat responses in Action 1; Capabilities: Browser navigation, element interaction, and file uploading via automation tools; Sanitization: Pre-processing of synthesized content mentioned in Action 2.\n- [SAFE]: All included Python scripts are transparent and use only standard libraries for internal routing and template generation, with no external dependencies or obfuscated code.

notebooklm