The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests potentially untrusted application state or user content into the state field of the Jev AI API request.
Ingestion points: The agent is instructed to populate the state parameter with facts needed for a decision as described in SKILL.md.
Boundary markers: The skill includes instructions to "prepare minimal state" and "remove secrets," but lacks programmatic structural delimiters to isolate untrusted data from the decision-making context.
Capability inventory: The skill utilizes curl to make network requests to the Jev AI service.
Sanitization: Instructions recommend the manual removal of sensitive data, but no automated filtering or sanitization is implemented.
[COMMAND_EXECUTION]: The skill provides instructions for the agent to use the curl command to communicate with the REST API at thejevai.com.
[EXTERNAL_DOWNLOADS]: The installation documentation recommends using npx to fetch the skill package from the vendor's repository.

jev-agent