The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection due to its core function of processing external data alongside powerful execution tools.
Ingestion points: The skill is designed to read example_dataset.csv and other user-supplied data files mentioned in SKILL.md for analysis and model training.
Boundary markers: The instructions do not specify any delimiters or safety prompts to prevent the agent from following instructions that might be hidden within the input datasets.
Capability inventory: The agent is authorized to use the Bash tool with unrestricted command execution (cmd:*), as well as Write and Edit tools, which could lead to unauthorized system actions if the agent is manipulated by malicious data.
Sanitization: There are no documented mechanisms for sanitizing or validating the content of the ingested data files before they are processed by the agent.

training-machine-learning-models