mercadona-cli-shopping
Fail
Audited by Gen Agent Trust Hub on Jul 3, 2026
Risk Level: CRITICALREMOTE_CODE_EXECUTIONCREDENTIALS_UNSAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The skill provides installation instructions that download a shell script from an external repository (
https://raw.githubusercontent.com/ivorpad/mercadona-cli/main/install.sh) and pipe it directly into the shell (sh). This pattern of piped remote execution is highly dangerous as it allows arbitrary code to run on the user's system without prior inspection. - [CREDENTIALS_UNSAFE]: The skill includes commands like
import-harandimport-curlthat ingest sensitive browser session exports. HAR files often contain full session cookies and authentication headers for all active sites, which can lead to the theft of the user's identity and account credentials if mishandled by the agent or the CLI tool. - [COMMAND_EXECUTION]: The skill enables the agent to use the
checkout submit --yescommand. This flag explicitly bypasses the confirmation step for placing real-world grocery orders, allowing the agent to commit the user to financial transactions autonomously. - [EXTERNAL_DOWNLOADS]: The skill relies on a third-party scoped NPM package (
@ivorpad/mercadona) and a Go binary from a personal GitHub repository. These sources are not verified or managed by a recognized trusted organization, increasing the risk of supply chain attacks. - [PROMPT_INJECTION]: The skill processes untrusted external data files (e.g., shopping lists, basket files) to determine which items to purchase. An attacker could use these files to inject malicious instructions that manipulate the quantities or types of products ordered.
- Ingestion points: File inputs for
batch -f,total -f, andcart set-many -fcommands. - Boundary markers: Absent; the skill does not use delimiters or instructions to prevent the agent from obeying embedded directives in the data.
- Capability inventory: Capabilities include modifying the shopping cart and submitting final orders.
- Sanitization: The skill includes budget guards (
--max), but these only limit the total price and do not sanitize the specific product IDs or instructions provided in the input files.
Recommendations
- HIGH: Downloads and executes remote code from: https://raw.githubusercontent.com/ivorpad/mercadona-cli/main/install.sh - DO NOT USE without thorough review
- AI detected serious security threats
Audit Metadata