The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes vehicle data from the Tessie API, which serves as a potential vector for indirect prompt injection if metadata (such as vehicle names) contains malicious instructions.
Ingestion points: Subcommands status, state, and vehicles in SKILL.md fetch external data.
Boundary markers: The skill does not define specific delimiters for separating vehicle metadata from instructions.
Capability inventory: The tessie-api command provides significant capabilities, including vehicle control and software update management.
Sanitization: The instructions do not specify sanitization routines for vehicle-returned strings.
[COMMAND_EXECUTION]: The skill's primary functionality is implemented through a local binary tessie-api, which is invoked for authentication, data retrieval, and command execution.
[SAFE]: Authentication is handled through a secure token proxy configuration and CLI helpers. The instructions explicitly prohibit the agent from displaying API tokens in the chat, reducing the risk of credential exposure.

tessie