The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The test script scripts/test_open_notebook_skill.py utilizes the Python compile() function to validate the syntax of the included example scripts. While used here for automated testing of the skill's own components, dynamic compilation of strings is a pattern that can be leveraged for arbitrary code execution if inputs are not strictly controlled.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface through its primary functionality:
Ingestion points: External data enters the context via web URLs and file uploads through the /api/sources endpoint as described in SKILL.md and references/api_reference.md.
Boundary markers: Instructions do not specify the use of delimiters or 'ignore' directives to prevent the agent from obeying instructions embedded within processed documents.
Capability inventory: The skill enables several high-level capabilities, including context-aware chat (scripts/chat_interaction.py) and content transformations, which are triggered based on the ingested data.
Sanitization: No validation or sanitization of the content from external sources is described before it is passed to the AI models.
[EXTERNAL_DOWNLOADS]: The SKILL.md file contains instructions to download a docker-compose.yml configuration file from a third-party repository (github.com/lfnovo) using curl. While standard for this software's installation, users should verify the contents of remote configuration files.
[COMMAND_EXECUTION]: The skill documentation includes instructions for the agent or user to execute local shell commands for system configuration and service deployment, including docker-compose operations and environment variable exports.

open-notebook