The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill relies on an external Python package named jev-skill and references assets and documentation from the author's GitHub repository (github.com/wuyoscar/jev-skill). These resources are hosted by the skill's developer and are required for the skill's primary function.
[COMMAND_EXECUTION]: The skill uses the jev-decide command-line utility to perform its core logic. The instructions explicitly direct the agent not to install dependencies automatically, but rather to inform the user if they are missing, ensuring user oversight over the environment.
[INDIRECT_PROMPT_INJECTION]: The skill has a surface for indirect prompt injection as it processes external JSON data to drive decision-making logic.
Ingestion points: External JSON files containing world state and decision criteria (e.g., assets/example.json) are passed to the jev-decide tool.
Boundary markers: The skill uses a structured JSON schema to separate state information from instructions and criteria, providing a degree of separation between data and instructions.
Capability inventory: The skill executes shell commands (jev-decide) and performs network requests to specified AI providers (OpenRouter or TypeSafe) to get decisions.
Sanitization: The skill utilizes a --dry-run flag for validation and requires the agent to review input data and obtain authorization before making live API calls.

jev-simulation