fiftyone-create-notebook

Fail

Audited by Gen Agent Trust Hub on Sep 14, 2026

Risk Level: HIGHREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The README.md file includes an installation instruction that pipes a remote script directly into the shell using curl -sL skil.sh | sh. Additionally, the generated notebook templates include code cells that perform !pip install operations, which execute remote code installation at runtime.
  • [DYNAMIC_EXECUTION]: The validation phase in SKILL.md instructs the agent to execute the newly generated notebook file using papermill. This involves the execution of code that was dynamically created by the agent based on user input, which is a significant execution risk.
  • [COMMAND_EXECUTION]: The skill requires the agent to execute multiple shell commands during the 'Validation' phase, including python -m venv, pip install, ipykernel install, papermill, and rm -rf to manage test environments and run notebook content.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes natural language requests from users to generate executable code for notebooks. This creates a surface where a malicious user request could influence the generated code cells, potentially leading to the inclusion of harmful commands that are then executed during the skill's validation step.
  • [EXTERNAL_DOWNLOADS]: The skill is configured to fetch technical documentation from https://docs.voxel51.com/llms.txt using a web retrieval tool to ensure API consistency. This is a vendor-owned resource used for retrieving current SDK patterns.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Sep 14, 2026, 02:54 PM