The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill's installation instructions download and execute code from external sources. It utilizes Homebrew to install from the 'steipete/tap/ordercli' repository and the Go toolchain to install from 'github.com/steipete/ordercli'.
[CREDENTIALS_UNSAFE]: The tool is designed to manage sensitive user authentication data. It accepts passwords via standard input, imports cookies from local Chrome browser profiles, and requires environment variables such as 'DELIVEROO_BEARER_TOKEN' for authentication.
[COMMAND_EXECUTION]: The skill operates by executing an external binary ('ordercli') with various arguments. This binary is downloaded at runtime and performs network operations to interact with private food delivery APIs.
[PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection by processing untrusted data from external sources. * Ingestion points: External data enters the agent context through commands that retrieve order history ('ordercli foodora history') and active order statuses ('ordercli foodora orders'). * Boundary markers: There are no specific delimiters or instructions provided to the agent to treat data from these external sources as untrusted or to ignore embedded instructions. * Capability inventory: The skill has the capability to modify account state, specifically through reordering actions ('ordercli foodora reorder --confirm') and session management. * Sanitization: No sanitization, validation, or filtering of the content retrieved from the delivery platforms is specified before it is handled by the agent.

ordercli