The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill provides an installation command that downloads a script from an untrusted third-party GitHub repository (ivorpad/mercadona-cli) and pipes it directly into the shell (curl | sh). This pattern is a critical security risk as it allows for arbitrary code execution from a source that is not recognized as a trusted organization or well-known service.
[EXTERNAL_DOWNLOADS]: The skill recommends installing a global Node.js package (@ivorpad/mercadona) and cloning a Go repository from an unverified developer. These external dependencies represent a supply chain risk as their integrity and maintenance are not guaranteed by a trusted vendor.
[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests untrusted data from external sources (like recipes or shopping lists) without boundary markers or sanitization. This is particularly dangerous because the skill has the capability to perform real-world financial transactions through its checkout submission functionality. 1. Ingestion points: Data passed to the batch, total, and search commands. 2. Boundary markers: Absent. 3. Capability inventory: The skill can access network APIs, write local configuration files, and place irreversible supermarket orders. 4. Sanitization: No evidence of validation for quantities or identifiers derived from untrusted input.
[COMMAND_EXECUTION]: The skill utilizes shell commands to interact with a CLI tool that handles sensitive authentication tokens and executes financial orders. While budget guards are mentioned, they are local configurations that do not fully mitigate the risk of an agent being manipulated into performing unauthorized actions.

mercadona-cli-shopping