The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions facilitate interaction with the local openydt CLI binary to perform parking-related operations. These commands are restricted to the parking domain (e.g., openydt parking get-car-in-list, openydt parking lock-car) and include explicit requirements for user confirmation (--yes flags) for write operations.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface typical of data-management tools, where output from read commands is used as input for write commands. This is documented as a functional requirement for the business flow.
Ingestion points: Data is ingested through query commands like get-park-on-site-car, get-car-in-list, and get-car-arrearage-list as described in SKILL.md and references/flows.md.
Boundary markers: The instructions lack explicit delimiters for variable interpolation, although they mandate using output values (like parkingCode or recordId) from read commands as inputs for subsequent actions.
Capability inventory: The skill has the capability to modify records, lock/unlock vehicles, and cancel arrears using the openydt CLI tool.
Sanitization: No specific sanitization or validation steps are outlined for the data returned by the CLI before it is processed by the agent.

openydt-record