The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes dataset examples which are interpolated into prompts in the infer.py template, creating a surface for indirect prompt injection.
Ingestion points: SKILL.md (Workflows section) and the exported dataset JSON files.
Boundary markers: The script template lacks delimiters or instructions for the model to ignore instructions within the dataset fields.
Capability inventory: The agent can execute ax CLI commands and Python scripts.
Sanitization: No sanitization is performed on the data ingested from the datasets.
[COMMAND_EXECUTION]: The skill suggests persisting configuration by appending environment variable exports to shell profile files (~/.zshrc, ~/.bashrc) using shell commands.
[EXTERNAL_DOWNLOADS]: The documentation includes instructions for installing several well-known packages, such as arize-ax-cli, openai, anthropic, and google-genai, which are standard for the skill's domain.

arize-experiment